Use of Latent Class Analysis and k-Means Clustering to Identify Complex Patient Profiles

Machine learning-based cluster analysis identifies four unique phenotypes of patients with degenerative cervical myelopathy with distinct clinical profiles and long-term functional and neurological outcomes

BACKGROUND

Degenerative cervical myelopathy (DCM), the predominant cause of spinal cord dysfunction among adults, exhibits diverse interrelated symptoms and significant heterogeneity in clinical presentation. This study sought to use machine learning-based clustering algorithms to identify distinct patient clinical profiles and functional trajectories following surgical intervention.

METHODS

In this study, we applied k-means and latent profile analysis (LPA) to identify patient phenotypes, using aggregated data from three major DCM trials. The combination of Nurick score, NDI (neck disability index), neck pain, as well as motor and sensory scores facilitated clustering. Goodness-of-fit indices were used to determine the optimal cluster number. ANOVA and post hoc Tukey's test assessed outcome differences, while multinomial logistic regression identified significant predictors of group membership.

FINDINGS

A total of 1047 patients with DCM (mean [SD] age: 56.80 [11.39] years, 411 [39%] females) had complete one year outcome assessment post-surgery. Latent profile analysis identified four DCM phenotypes: "severe multimodal impairment" (n = 286), "minimal impairment" (n = 116), "motor-dominant" (n = 88) and "pain-dominant" (n = 557) groups. Each phenotype exhibited a unique symptom profile and distinct functional recovery trajectories. The "severe multimodal impairment group", comprising frail elderly patients, demonstrated the worst overall outcomes at one year (SF-36 PCS mean [SD]: 40.01 [9.75]; SF-36 MCS mean [SD], 46.08 [11.50]) but experienced substantial neurological recovery post-surgery (ΔmJOA mean [SD]: 3.83 [2.98]). Applying the k-means algorithm yielded a similar four-class solution. A higher frailty score and positive smoking status predicted membership in the "severe multimodal impairment" group (OR 1.47 [95% CI 1.07-2.02] and 1.58 [95% CI 1.25-1.99, respectively]), while undergoing anterior surgery and a longer symptom duration were associated with the "pain-dominant" group (OR 2.0 [95% CI 1.06-3.80] and 3.1 [95% CI 1.38-6.89], respectively).

INTERPRETATION

Unsupervised learning on multiple clinical metrics predicted distinct patient phenotypes. Symptom clustering offers a valuable framework to identify DCM subpopulations, surpassing single patient reported outcome measures like the mJOA.

FUNDING

No funding was received for the present work. The original studies were funded by AO Spine North America.

Depressive symptoms and neuroticism mediate the association between traumatic events and suicidality - A latent class mediation analysis of UK Biobank Database

BACKGROUND

Mental disorders that follow traumatic experience may increase risk of suicidality, but a comprehensive approach to understand how these mental disorders mediate the association between psychological traumatic experience and suicidality should be elucidated. In this study, we attempted to provide comprehensive evidence on how depressive symptoms and neuroticism can mediate the association between psychological traumatic experiences and suicidal behaviours including suicidal ideation, suicidal planning, and suicide attempts.

METHODS

We analyzed 111,931 participants from UK Biobank who had completed mental health web-based questionnaire from 2016 to 2017. "Self-harm and suicidal behaviour and ideation (SSBI) score" was calculated by the response from suicidal behaviours and self-harm questionnaires. Conducting multivariate linear regression, depressive symptoms, anxiety symptoms, and neuroticism were selected as potential mediators. We constructed a latent class mediation model estimated direct effect of psychological traumatic events on suicidality and indirect effect of psychological traumatic events mediated by depressive symptoms and neuroticism.

RESULTS

Psychological traumatic events were positively associated with suicidal behaviours. Depressive symptoms and neuroticism significantly mediated the effect of psychological traumatic events on suicidality. Anxiety symptoms did not mediate the association between psychological traumatic events and suicidality.

CONCLUSION

Psychological traumatic events, irrespective of life stage of occurrence, are associated with suicidality. The association between psychological traumatic events and suicidality can be partially explained by depressive symptoms and neuroticism of those who were exposed to psychological trauma.

Novel Preoperative Risk Stratification Using Digital Phenotyping Applying a Scalable Machine-Learning Approach

BACKGROUND

Classification of perioperative risk is important for patient care, resource allocation, and guiding shared decision-making. Using discriminative features from the electronic health record (EHR), machine-learning algorithms can create digital phenotypes among heterogenous populations, representing distinct patient subpopulations grouped by shared characteristics, from which we can personalize care, anticipate clinical care trajectories, and explore therapies. We hypothesized that digital phenotypes in preoperative settings are associated with postoperative adverse events including in-hospital and 30-day mortality, 30-day surgical redo, intensive care unit (ICU) admission, and hospital length of stay (LOS).

METHODS

We identified all laminectomies, colectomies, and thoracic surgeries performed over a 9-year period from a large hospital system. Seventy-seven readily extractable preoperative features were first selected from clinical consensus, including demographics, medical history, and lab results. Three surgery-specific datasets were built and split into derivation and validation cohorts using chronological occurrence. Consensus k -means clustering was performed independently on each derivation cohort, from which phenotypes' characteristics were explored. Cluster assignments were used to train a random forest model to assign patient phenotypes in validation cohorts. We reconducted descriptive analyses on validation cohorts to confirm the similarity of patient characteristics with derivation cohorts, and quantified the association of each phenotype with postoperative adverse events by using the area under receiver operating characteristic curve (AUROC). We compared our approach to American Society of Anesthesiologists (ASA) alone and investigated a combination of our phenotypes with the ASA score.

RESULTS

A total of 7251 patients met inclusion criteria, of which 2770 were held out in a validation dataset based on chronological occurrence. Using segmentation metrics and clinical consensus, 3 distinct phenotypes were created for each surgery. The main features used for segmentation included urgency of the procedure, preoperative LOS, age, and comorbidities. The most relevant characteristics varied for each of the 3 surgeries. Low-risk phenotype alpha was the most common (2039 of 2770, 74%), while high-risk phenotype gamma was the rarest (302 of 2770, 11%). Adverse outcomes progressively increased from phenotypes alpha to gamma, including 30-day mortality (0.3%, 2.1%, and 6.0%, respectively), in-hospital mortality (0.2%, 2.3%, and 7.3%), and prolonged hospital LOS (3.4%, 22.1%, and 25.8%). When combined with the ASA score, digital phenotypes achieved higher AUROC than the ASA score alone (hospital mortality: 0.91 vs 0.84; prolonged hospitalization: 0.80 vs 0.71).

CONCLUSIONS

For 3 frequently performed surgeries, we identified 3 digital phenotypes. The typical profiles of each phenotype were described and could be used to anticipate adverse postoperative events.

Predicting Self-Reported Social Risk in Medically Complex Adults Using Electronic Health Data

BACKGROUND

Social barriers to health care, such as food insecurity, financial distress, and housing instability, may impede effective clinical management for individuals with chronic illness. Systematic strategies are needed to more efficiently identify at-risk individuals who may benefit from proactive outreach by health care systems for screening and referral to available social resources.

OBJECTIVE

To create a predictive model to identify a higher likelihood of food insecurity, financial distress, and/or housing instability among adults with multiple chronic medical conditions.

RESEARCH DESIGN AND SUBJECTS

We developed and validated a predictive model in adults with 2 or more chronic conditions who were receiving care within Kaiser Permanente Northern California (KPNC) between January 2017 and February 2020. The model was developed to predict the likelihood of a "yes" response to any of 3 validated self-reported survey questions related to current concerns about food insecurity, financial distress, and/or housing instability. External model validation was conducted in a separate cohort of adult non-Medicaid KPNC members aged 35-85 who completed a survey administered to a random sample of health plan members between April and June 2021 (n = 2820).

MEASURES

We examined the performance of multiple model iterations by comparing areas under the receiver operating characteristic curves (AUCs). We also assessed algorithmic bias related to race/ethnicity and calculated model performance at defined risk thresholds for screening implementation.

RESULTS

Patients in the primary modeling cohort (n = 11,999) had a mean age of 53.8 (±19.3) years, 64.7% were women, and 63.9% were of non-White race/ethnicity. The final, simplified model with 30 predictors (including utilization, diagnosis, behavior, insurance, neighborhood, and pharmacy-based variables) had an AUC of 0.68. The model remained robust within different race/ethnic strata.

CONCLUSIONS

Our results demonstrated that a predictive model developed using information gleaned from the medical record and from public census tract data can be used to identify patients who may benefit from proactive social needs assessment. Depending on the prevalence of social needs in the target population, different risk output thresholds could be set to optimize positive predictive value for successful outreach. This predictive model-based strategy provides a pathway for prioritizing more intensive social risk outreach and screening efforts to the patients who may be in greatest need.

Decoding Suicide Decedent Profiles and Signs of Suicidal Intent Using Latent Class Analysis

Importance

Suicide rates in the US increased by 35.6% from 2001 to 2021. Given that most individuals die on their first attempt, earlier detection and intervention are crucial. Understanding modifiable risk factors is key to effective prevention strategies.

Objective

To identify distinct suicide profiles or classes, associated signs of suicidal intent, and patterns of modifiable risks for targeted prevention efforts.

Design, Setting, and Participants

This cross-sectional study used data from the 2003-2020 National Violent Death Reporting System Restricted Access Database for 306 800 suicide decedents. Statistical analysis was performed from July 2022 to June 2023.

Exposures

Suicide decedent profiles were determined using latent class analyses of available data on suicide circumstances, toxicology, and methods.

Main Outcomes and Measures

Disclosure of recent intent, suicide note presence, and known psychotropic usage.

Results

Among 306 800 suicide decedents (mean [SD] age, 46.3 [18.4] years; 239 627 males [78.1%] and 67 108 females [21.9%]), 5 profiles or classes were identified. The largest class, class 4 (97 175 [31.7%]), predominantly faced physical health challenges, followed by polysubstance problems in class 5 (58 803 [19.2%]), and crisis, alcohol-related, and intimate partner problems in class 3 (55 367 [18.0%]), mental health problems (class 2, 53 928 [17.6%]), and comorbid mental health and substance use disorders (class 1, 41 527 [13.5%]). Class 4 had the lowest rates of disclosing suicidal intent (13 952 [14.4%]) and leaving a suicide note (24 351 [25.1%]). Adjusting for covariates, compared with class 1, class 4 had the highest odds of not disclosing suicide intent (odds ratio [OR], 2.58; 95% CI, 2.51-2.66) and not leaving a suicide note (OR, 1.45; 95% CI, 1.41-1.49). Class 4 also had the lowest rates of all known psychiatric illnesses and psychotropic medications among all suicide profiles. Class 4 had more older adults (23 794 were aged 55-70 years [24.5%]; 20 100 aged ≥71 years [20.7%]), veterans (22 220 [22.9%]), widows (8633 [8.9%]), individuals with less than high school education (15 690 [16.1%]), and rural residents (23 966 [24.7%]).

Conclusions and Relevance

This study identified 5 distinct suicide profiles, highlighting a need for tailored prevention strategies. Improving the detection and treatment of coexisting mental health conditions, substance and alcohol use disorders, and physical illnesses is paramount. The implementation of means restriction strategies plays a vital role in reducing suicide risks across most of the profiles, reinforcing the need for a multifaceted approach to suicide prevention.

Reading tea leaves or tracking true constructs? An assessment of personality-based latent profiles in eating disorders

Background

Eating disorder (ED) subtyping studies have often extracted an undercontrolled, an overcontrolled and a resilient profile based on trait impulsivity and perfectionism. However, the extent to which methodological choices impact the coherence and distinctness of resulting subtypes remains unclear.

Objective

In this paper, we aimed to assess the robustness of these findings by extracting personality-based subtypes on a sample of ED patients (N = 221) under different analytic conditions.

Methods

We ran four latent profile analyses (LPA), varying the extent to which we constrained variances and covariances during model parametrization. We then performed a comparative analysis also including state ED symptom measures as indicators. Finally, we used cross-method validation via k-means clustering to further assess the robustness of our profiles.

Results

Our results demonstrated a four-profile model based on variances in impulsivity and perfectionism to fit the data well. Across model solutions, the profiles with the most and least state and trait disturbances were replicated most stably, while more nuanced variations in trait variables resulted in less consistent profiles. Inclusion of ED symptoms as indicator variables increased subtype differentiation and similarity across profiles. Validation cluster analyses aligned most with more restrictive LPA models.

Conclusion

These results suggest that ED subtypes track true constructs, since subtypes emerged method-independently. We found analytic methods to constrain the theoretical and practical conclusions that can be drawn. This underscores the importance of objective-driven analytic design and highlights its relevance in applying research findings in clinical practice.

COVID-19 patient profiles over four waves in Barcelona metropolitan area: A clustering approach

OBJECTIVES

Identifying profiles of hospitalized COVID-19 patients and explore their association with different degrees of severity of COVID-19 outcomes (i.e. in-hospital mortality, ICU assistance, and invasive mechanical ventilation). The findings of this study could inform the development of multiple care intervention strategies to improve patient outcomes.

METHODS

Prospective multicentre cohort study during four different waves of COVID-19 from March 1st, 2020 to August 31st, 2021 in four health consortiums within the southern Barcelona metropolitan region. From a starting point of over 292 demographic characteristics, comorbidities, vital signs, severity scores, and clinical analytics at hospital admission, we used both clinical judgment and supervised statistical methods to reduce to the 36 most informative completed covariates according to the disease outcomes for each wave. Patients were then grouped using an unsupervised semiparametric method (KAMILA). Results were interpreted by clinical and statistician team consensus to identify clinically-meaningful patient profiles.

RESULTS

The analysis included nw1 = 1657, nw2 = 697, nw3 = 677, and nw4 = 787 hospitalized-COVID-19 patients for each of the four waves. Clustering analysis identified 2 patient profiles for waves 1 and 3, while 3 profiles were determined for waves 2 and 4. Patients allocated in those groups showed a different percentage of disease outcomes (e.g., wave 1: 15.9% (Cluster 1) vs. 31.8% (Cluster 2) for in-hospital mortality rate). The main factors to determine groups were the patient's age and number of obese patients, number of comorbidities, oxygen support requirement, and various severity scores. The last wave is also influenced by the massive incorporation of COVID-19 vaccines.

CONCLUSION

Our study suggests that a single care model at hospital admission may not meet the needs of hospitalized-COVID-19 adults. A clustering approach appears to be appropriate for helping physicians to differentiate patients and, thus, apply multiple care intervention strategies, as another way of responding to new outbreaks of this or future diseases.

The influence of attributes on community preferences regarding antibiotic treatment: evidence from a discrete choice model

Antibiotic resistance (AR) rates in Vietnam are among the highest in Asia, and recent infections due to multi-drug resistance in the country have caused thousands of deaths each year. This study investigated a Vietnamese community's preferences for antibiotic treatment and its knowledge and attitudes regarding antibiotics. A discrete choice experiment-based survey was developed and administered to the population of interest. The respondents were given sociodemographic-, knowledge- and attitude-related items and 17 pairs of choice tasks. Two hypothetical options were included in each choice task. Latent class analysis was conducted to determine the differences among the respondents' preferences. Among 1,014 respondents, 805 (79.4%) gave valid questionnaires. A three-latent-class model with four covariates (age, healthcare-related education or career, occupation, and attitude classifications) was used in the analysis. All five attributes significantly influenced the respondents' decisions. The majority, including young employed respondents with non-healthcare-related work or education, found treatment failure more important. Older respondents who had healthcare-related education/careers and/or appropriate antibiotic use- and antibiotics resistance-related attitudes, regarded contribution to antibiotic resistance as an important attribute in selecting antibiotic treatments. Unemployed individuals with correct knowledge identified the cost of antibiotic treatment as the most essential decision-making factor. Findings suggest minimal antibiotic impact on resistance; only 7.83% view it as amajor concern. The respondents exhibited substantial preference heterogeneity, and the general Vietnamese public had poor knowledge of and attitudes toward antibiotic use and antibiotic resistance. This study emphasizes the need for individual responsibility for antibiotic resistance and appropriate antibiotic use.

Phenotyping Down syndrome: discovery and predictive modelling with electronic medical records

BACKGROUND

Individuals with Down syndrome (DS) have a heightened risk for various co-occurring health conditions, including congenital heart disease (CHD). In this two-part study, electronic medical records (EMRs) were leveraged to examine co-occurring health conditions among individuals with DS (Study 1) and to investigate health conditions linked to surgical intervention among DS cases with CHD (Study 2).

METHODS

De-identified EMRs were acquired from Vanderbilt University Medical Center and facilitated creating a cohort of N = 2282 DS cases (55% females), along with comparison groups for each study. In Study 1, DS cases were one-by-two sex and age matched with samples of case-controls and of individuals with other intellectual and developmental difficulties (IDDs). The phenome-disease association study (PheDAS) strategy was employed to reveal co-occurring health conditions in DS versus comparison groups, which were then ranked for how often they are discussed in relation to DS using the PubMed database and Novelty Finding Index. In Study 2, a subset of DS individuals with CHD [N = 1098 (48%)] were identified to create longitudinal data for N = 204 cases with surgical intervention (19%) versus 204 case-controls. Data were included in predictive models and assessed which model-based health conditions, when more prevalent, would increase the likelihood of surgical intervention.

RESULTS

In Study 1, relative to case-controls and those with other IDDs, co-occurring health conditions among individuals with DS were confirmed to include heart failure, pulmonary heart disease, atrioventricular block, heart transplant/surgery and primary pulmonary hypertension (circulatory); hypothyroidism (endocrine/metabolic); and speech and language disorder and Alzheimer's disease (neurological/mental). Findings also revealed more versus less prevalent co-occurring health conditions in individuals with DS when comparing with those with other IDDs. Findings with high Novelty Finding Index were abnormal electrocardiogram, non-rheumatic aortic valve disorders and heart failure (circulatory); acid-base balance disorder (endocrine/metabolism); and abnormal blood chemistry (symptoms). In Study 2, the predictive models revealed that among individuals with DS and CHD, presence of health conditions such as congestive heart failure (circulatory), valvular heart disease and cardiac shunt (congenital), and pleural effusion and pulmonary collapse (respiratory) were associated with increased likelihood of surgical intervention.

CONCLUSIONS

Research efforts using EMRs and rigorous statistical methods could shed light on the complexity in health profile among individuals with DS and other IDDs and motivate precision-care development.

Psychosocial adaptation profiles in young and middle-aged patients with acute myocardial infarction: a latent profile analysis

AIMS

We sought to explore the latent classifications of psychosocial adaptation in young and middle-aged patients with acute myocardial infarction (AMI) and analyse the characteristics of different profiles of AMI patients.

METHODS AND RESULTS

A cross-sectional study was performed in 438 Chinese young and middle-aged patients with AMI. The investigation time was 1 month after discharge. Three different self-report instruments were distributed to the participants, including the Psychosocial Adjustment to Illness Scale, the Perceived Stress Scale, and the Social Support Rating Scale. The seven dimensions of the Psychosocial Adjustment to Illness Scale were then used to perform a latent profile analysis. All participants signed informed consent forms in accordance with the ethical principles of the Declaration of Helsinki. Finally, a total of 411 young and middle-aged AMI patients were enrolled. Three distinct profiles were identified, including the 'well-adapted group' (44.8%), 'highlight in psychological burdens group' (25.5%), and 'poorly adapted group' (29.7%). The influencing factors included stress perception, social support, occupational type, and marital status (P < 0.05).

CONCLUSION

The psychosocial adaptation of young and middle-aged AMI patients can be divided into three profiles. Clinical nurses can carry out individualized psychological interventions according to the characteristics of patients in different potential profiles to improve the psychosocial adaptation of patients and the prognosis of their disease.

Association of socioeconomic deprivation with outcomes in critically ill adult patients: an observational prospective multicenter cohort study

BACKGROUND

The influence of socioeconomic deprivation on health inequalities is established, but its effect on critically ill patients remains unclear, due to inconsistent definitions in previous studies.

METHODS

Prospective multicenter cohort study conducted from March to June 2018 in eight ICUs in the Greater Paris area. All admitted patients aged ≥ 18 years were enrolled. Socioeconomic phenotypes were identified using hierarchical clustering, based on education, health insurance, income, and housing. Association of phenotypes with 180-day mortality was assessed using Cox proportional hazards models.

RESULTS

A total of 1,748 patients were included. Median age was 62.9 [47.4-74.5] years, 654 (37.4%) patients were female, and median SOFA score was 3 [1-6]. Study population was clustered in five phenotypes with increasing socioeconomic deprivation. Patients from phenotype A (n = 958/1,748, 54.8%) were without socioeconomic deprivation, patients from phenotype B (n = 273/1,748, 15.6%) had only lower education levels, phenotype C patients (n = 117/1,748, 6.7%) had a cumulative burden of 1[1-2] deprivations and all had housing deprivation, phenotype D patients had 2 [1-2] deprivations, all of them with income deprivation, and phenotype E patients (n = 93/1,748, 5.3%) included patients with 3 [2-4] deprivations and included all patients with health insurance deprivation. Patients from phenotypes D and E were younger, had fewer comorbidities, more alcohol and opiate use, and were more frequently admitted due to self-harm diagnoses. Patients from phenotype C (predominant housing deprivation), were more frequently admitted with diagnoses related to chronic respiratory diseases and received more non-invasive positive pressure ventilation. Following adjustment for age, sex, alcohol and opiate use, socioeconomic phenotypes were not associated with increased 180-day mortality: phenotype A (reference); phenotype B (hazard ratio [HR], 0.85; 95% confidence interval CI 0.65-1.12); phenotype C (HR, 0.56; 95% CI 0.34-0.93); phenotype D (HR, 1.09; 95% CI 0.78-1.51); phenotype E (HR, 1.20; 95% CI 0.73-1.96).

CONCLUSIONS

In a universal health care system, the most deprived socioeconomic phenotypes were not associated with increased 180-day mortality. The most disadvantaged populations exhibit distinct characteristics and medical conditions that may be addressed through targeted public health interventions.

Contrasting Objective and Perceived Risk: Predicting COVID-19 Health Behaviors in a Nationally Representative U.S. Sample

BACKGROUND

Individuals confronting health threats may display an optimistic bias such that judgments of their risk for illness or death are unrealistically positive given their objective circumstances.

PURPOSE

We explored optimistic bias for health risks using k-means clustering in the context of COVID-19. We identified risk profiles using subjective and objective indicators of severity and susceptibility risk for COVID-19.

METHODS

Between 3/18/2020-4/18/2020, a national probability sample of 6,514 U.S. residents reported both their subjective risk perceptions (e.g., perceived likelihood of illness or death) and objective risk indices (e.g., age, weight, pre-existing conditions) of COVID-19-related susceptibility and severity, alongside other pandemic-related experiences. Six months later, a subsample (N = 5,661) completed a follow-up survey with questions about their frequency of engagement in recommended health protective behaviors (social distancing, mask wearing, risk behaviors, vaccination intentions).

RESULTS

The k-means clustering procedure identified five risk profiles in the Wave 1 sample; two of these demonstrated aspects of optimistic bias, representing almost 44% of the sample. In OLS regression models predicting health protective behavior adoption at Wave 2, clusters representing individuals with high perceived severity risk were most likely to report engagement in social distancing, but many individuals who were objectively at high risk for illness and death did not report engaging in self-protective behaviors.

CONCLUSIONS

Objective risk of disease severity only inconsistently predicted health protective behavior. Risk profiles may help identify groups that need more targeted interventions to increase their support for public health policy and health enhancing recommendations more broadly.

Association between systemic immune inflammation index, systemic inflammation response index and adult psoriasis: evidence from NHANES

Background

The systemic immune-inflammation index (SII) and systemic inflammation response index (SIRI) are both novel biomarkers and predictors of inflammation. Psoriasis is a skin disease characterized by chronic inflammation. This study aimed to investigate the potential association between SII, SIRI, and adult psoriasis.

Methods

Data of adults aged 20 to 80 years from the National Health and Nutrition Examination Survey (NHANES) (2003-2006, 2009-2014) were utilized. The K-means method was used to group SII and SIRI into low, medium, and high-level clusters. Additionally, SII or SIRI levels were categorized into three groups: low (1st-3rd quintiles), medium (4th quintile), and high (5th quintile). The association between SII-SIRI pattern, SII or SIRI individually, and psoriasis was assessed using multivariate logistic regression models. The results were presented as odds ratios (ORs) and confidence intervals (CIs). Restricted cubic spline (RCS) regression, subgroup, and interaction analyses were also conducted to explore the potential non-linear and independent relationships between natural log-transformed SII (lnSII) levels or SIRI levels and psoriasis, respectively.

Results

Of the 18208 adults included in the study, 511 (2.81%) were diagnosed with psoriasis. Compared to the low-level group of the SII-SIRI pattern, participants in the medium-level group had a significantly higher risk for psoriasis (OR = 1.40, 95% CI: 1.09, 1.81, p-trend = 0.0031). In the analysis of SII or SIRI individually, both SII and SIRI were found to be positively associated with the risk of psoriasis (high vs. low group OR = 1.52, 95% CI: 1.18, 1.95, p-trend = 0.0014; OR = 1.48, 95% CI: 1.12, 1.95, p-trend = 0.007, respectively). Non-linear relationships were observed between lnSII/SIRI and psoriasis (both p-values for overall < 0.05, p-values for nonlinearity < 0.05). The association between SII levels and psoriasis was stronger in females, obese individuals, people with type 2 diabetes, and those without hypercholesterolemia.

Conclusion

We observed positive associations between SII-SIRI pattern, SII, SIRI, and psoriasis among U.S. adults. Further well-designed studies are needed to gain a better understanding of these findings.

Long COVID is not a uniform syndrome: Evidence from person-level symptom clusters using latent class analysis

BACKGROUND

The current study aims to enhance insight into the heterogeneity of long COVID by identifying symptom clusters and associated socio-demographic and health determinants.

METHODS

A total of 458 participants (Mage 36.0 ± 11.9; 46.5% male) with persistent symptoms after COVID-19 completed an online self-report questionnaire including a 114-item symptom list. First, a k-means clustering analysis was performed to investigate overall clustering patterns and identify symptoms that provided meaningful distinctions between clusters. Next, a step-three latent class analysis (LCA) was performed based on these distinctive symptoms to analyze person-centered clusters. Finally, multinominal logistic models were used to identify determinants associated with the symptom clusters.

RESULTS

From a 5-cluster solution obtained from k-means clustering, 30 distinctive symptoms were selected. Using LCA, six symptom classes were identified: moderate (20.7%) and high (20.7%) inflammatory symptoms, moderate malaise-neurocognitive symptoms (18.3%), high malaise-neurocognitive-psychosocial symptoms (17.0%), low-overall symptoms (13.3%) and high overall symptoms (9.8%). Sex, age, employment, COVID-19 suspicion, COVID-19 severity, number of acute COVID-19 symptoms, long COVID symptom duration, long COVID diagnosis, and impact of long COVID were associated with the different symptom clusters.

CONCLUSIONS

The current study's findings characterize the heterogeneity in long COVID symptoms and underscore the importance of identifying determinants of different symptom clusters.

A scoping review finds a growing trend in studies validating multimorbidity patterns and identifies five broad types of validation methods

OBJECTIVES

Multimorbidity, the presence of two or more long-term conditions, is a growing public health concern. Many studies use analytical methods to discover multimorbidity patterns from data. We aimed to review approaches used in published literature to validate these patterns.

STUDY DESIGN AND SETTING

We systematically searched PubMed and Web of Science for studies published between July 2017 and July 2023 that used analytical methods to discover multimorbidity patterns.

RESULTS

Out of 31,617 studies returned by the searches, 172 were included. Of these, 111 studies (64%) conducted validation, the number of studies with validation increased from 53.13% (17 out of 32 studies) to 71.25% (57 out of 80 studies) in 2017-2019 to 2022-2023, respectively. Five types of validation were identified: assessing the association of multimorbidity patterns with clinical outcomes (n = 79), stability across subsamples (n = 26), clinical plausibility (n = 22), stability across methods (n = 7) and exploring common determinants (n = 2). Some studies used multiple types of validation.

CONCLUSION

The number of studies conducting a validation of multimorbidity patterns is clearly increasing. The most popular validation approach is assessing the association of multimorbidity patterns with clinical outcomes. Methodological guidance on the validation of multimorbidity patterns is needed.

Neuro-ophthalmology Emergency Department and Inpatient Consultations at a Large Academic Referral Center

PURPOSE

Prompt neuro-ophthalmology consultation prevents diagnostic errors and improves patient outcomes. The scarcity of neuro-ophthalmologists means that the increasing outpatient demand cannot be met, prompting many emergency department (ED) referrals by non-neuro-ophthalmologists. We describe our quaternary care institution's ED and inpatient neuro-ophthalmology consultation patterns and patient outcomes.

DESIGN

Prospective observational study.

PARTICIPANTS

Consecutive neuro-ophthalmology ED and inpatient consultation requests over 1 year.

METHODS

We collected patient demographics, distance traveled, insurance status, referring provider details, consultation question, final diagnosis, complexity of consultation, time of consultation, and need for outpatient follow-up.

MAIN OUTCOME MEASURES

Consultation patterns and diagnoses, complexity, and follow-up.

RESULTS

Of 494 consecutive adult ED and inpatient neuro-ophthalmology consultations requested over 1 year, 241 of 494 consultations (49%) occurred at night or during weekends. Of ED consultations (322 of 494 [65%]), 127 of 322 consultations (39%) occurred during weekdays, 126 of 322 consultations (39%) occurred on weeknights, and 69 of 322 consultations (22%) occurred on weekends or holidays. Of 322 ED consultations, 225 of 322 consultations (70%) were patients who initially sought treatment in the ED with a neuro-ophthalmic chief symptom. Of the 196 patients sent to the ED by a health care professional, 148 patients (148/196 [76%]) were referred by eye care specialists (74 optometrists and 74 ophthalmologists). The most common ED referral questions were for papilledema (75 of 322 [23%]) and vision loss (72 of 322 [22%]). A total of 219 of 322 patients (68%) received a final active neuro-ophthalmic diagnosis, 222 of 322 patients (69%) were cases of high or very high complexity, and 143 of 322 patients (44%) required admission. Inpatient consultations (n = 172) were requested most frequently by hospitalists, including neurologists (71 of 172 [41%]) and oncologists (20 of 172 [12%]) for vision loss (43 of 172 [25%]) and eye movement disorders (36 of 172 [21%]) and by neurosurgeons (58 of 172 [33%]) for examination for mass or a preoperative evaluation (19 of 172 [11%]). An active neuro-ophthalmic diagnosis was confirmed in 67% of patients (116 of 172). Outpatient neuro-ophthalmology follow-up was required for 291 of 494 patients (59%).

CONCLUSIONS

Neuro-ophthalmology consultations are critical to the diagnosis and management in the hospital setting. In the face of a critical shortage of neuro-ophthalmologists, this study highlights the need for technological and diagnostic aids for greater outpatient access.

FINANCIAL DISCLOSURE(S)

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Characterizing hospitalization trajectories in the high-need, high-cost population using electronic health record data

High utilization by a minority of patients accounts for a large share of health care costs, but the dynamics of this utilization remain poorly understood. We sought to characterize longitudinal trajectories of hospitalization among adult patients at an academic medical center from 2017 to 2023. Among 3404 patients meeting eligibility criteria, following an initial "rising-risk" period of 3 hospitalizations in 6 months, growth mixture modeling discerned 4 clusters of subsequent hospitalization trajectories: no further utilization, low chronic utilization, persistently high utilization with a slow rate of increase, and persistently high utilization with a fast rate of increase. Baseline factors associated with higher-order hospitalization trajectories included admission to a nonsurgical service, full code status, intensive care unit-level care, opioid administration, discharge home, and comorbid cardiovascular disease, end-stage kidney or liver disease, or cancer. Characterizing hospitalization trajectories and their correlates in this manner lays groundwork for early identification of those most likely to become high-need, high-cost patients.

A comprehensive review of machine learning algorithms and their application in geriatric medicine: present and future

The increasing access to health data worldwide is driving a resurgence in machine learning research, including data-hungry deep learning algorithms. More computationally efficient algorithms now offer unique opportunities to enhance diagnosis, risk stratification, and individualised approaches to patient management. Such opportunities are particularly relevant for the management of older patients, a group that is characterised by complex multimorbidity patterns and significant interindividual variability in homeostatic capacity, organ function, and response to treatment. Clinical tools that utilise machine learning algorithms to determine the optimal choice of treatment are slowly gaining the necessary approval from governing bodies and being implemented into healthcare, with significant implications for virtually all medical disciplines during the next phase of digital medicine. Beyond obtaining regulatory approval, a crucial element in implementing these tools is the trust and support of the people that use them. In this context, an increased understanding by clinicians of artificial intelligence and machine learning algorithms provides an appreciation of the possible benefits, risks, and uncertainties, and improves the chances for successful adoption. This review provides a broad taxonomy of machine learning algorithms, followed by a more detailed description of each algorithm class, their purpose and capabilities, and examples of their applications, particularly in geriatric medicine. Additional focus is given on the clinical implications and challenges involved in relying on devices with reduced interpretability and the progress made in counteracting the latter via the development of explainable machine learning.

Health-related quality of life profiles in adolescents and young adults with chronic conditions

PURPOSE

To assess health-related quality of life (HRQOL) among adolescents and young adults (AYAs) with chronic conditions.

METHODS

AYAs (N = 872) aged 14-20 years completed NIH's Patient-Reported Outcomes Measurement Information System® (PROMIS®) measures of physical function, pain interference, fatigue, social health, depression, anxiety, and anger. Latent profile analysis (LPA) was used to group AYAs into HRQOL profiles using PROMIS T-scores. The optimal number of profiles was determined by model fit statistics, likelihood ratio test, and entropy. Multinomial logistic regression models were used to examine how LPA's HRQOL profile membership was associated with patient demographic and chronic conditions. The model prediction accuracy on profile membership was evaluated using Huberty's I index with a threshold of 0.35 for good effect.

RESULTS

A 4-profile LPA model was selected. A total of 161 (18.5%), 256 (29.4%), 364 (41.7%), and 91 (10.4%) AYAs were classified into Minimal, Mild, Moderate, and Severe HRQOL Impact profiles. AYAs in each profile had distinctive mean scores with over a half standard deviation (5-points in PROMIS T-scores) of difference between profiles across most HRQOL domains. AYAs who were female or had conditions such as mental health condition, hypertension, and self-reported chronic pain were more likely to be in the Severe HRQOL Impact profile. The Huberty's I index was 0.36.

CONCLUSIONS

Approximately half of AYAs with a chronic condition experience moderate to severe HRQOL impact. The availability of risk prediction models for HRQOL impact will help to identify AYAs who are in greatest need of closer clinical care follow-up.

Identification and Prediction of Clinical Phenotypes in Hospitalized Patients With COVID-19: Machine Learning From Medical Records

BACKGROUND

There is significant heterogeneity in disease progression among hospitalized patients with COVID-19. The pathogenesis of SARS-CoV-2 infection is attributed to a complex interplay between virus and host immune response that in some patients unpredictably and rapidly leads to "hyperinflammation" associated with increased risk of mortality. The early identification of patients at risk of progression to hyperinflammation may help inform timely therapeutic decisions and lead to improved outcomes.

OBJECTIVE

The primary objective of this study was to use machine learning to reproducibly identify specific risk-stratifying clinical phenotypes across hospitalized patients with COVID-19 and compare treatment response characteristics and outcomes. A secondary objective was to derive a predictive phenotype classification model using routinely available early encounter data that may be useful in informing optimal COVID-19 bedside clinical management.

METHODS

This was a retrospective analysis of electronic health record data of adult patients (N=4379) who were admitted to a Johns Hopkins Health System hospital for COVID-19 treatment from 2020 to 2021. Phenotypes were identified by clustering 38 routine clinical observations recorded during inpatient care. To examine the reproducibility and validity of the derived phenotypes, patient data were randomly divided into 2 cohorts, and clustering analysis was performed independently for each cohort. A predictive phenotype classifier using the gradient-boosting machine method was derived using routine clinical observations recorded during the first 6 hours following admission.

RESULTS

A total of 2 phenotypes (designated as phenotype 1 and phenotype 2) were identified in patients admitted for COVID-19 in both the training and validation cohorts with similar distributions of features, correlations with biomarkers, treatments, comorbidities, and outcomes. In both the training and validation cohorts, phenotype-2 patients were older; had elevated markers of inflammation; and were at an increased risk of requiring intensive care unit-level care, developing sepsis, and mortality compared with phenotype-1 patients. The gradient-boosting machine phenotype prediction model yielded an area under the curve of 0.89 and a positive predictive value of 0.83.

CONCLUSIONS

Using machine learning clustering, we identified and internally validated 2 clinical COVID-19 phenotypes with distinct treatment or response characteristics consistent with similar 2-phenotype models derived from other hospitalized populations with COVID-19, supporting the reliability and generalizability of these findings. COVID-19 phenotypes can be accurately identified using machine learning models based on readily available early encounter clinical data. A phenotype prediction model based on early encounter data may be clinically useful for timely bedside risk stratification and treatment personalization.

Telehealth During COVID-19 for Adults with Multiple Chronic Conditions: Associations with Self-Reported Food Insecurity and with Physical Limitations

Background: Adults with chronic medical conditions complicated by food insecurity or physical limitations may have higher barriers to accessing telehealth implemented during the COVID-19 pandemic. Objective: To examine the relationships of self-reported food insecurity and physical limitations with changes in health care utilization and medication adherence comparing the year before (March 2019-February 2020) and the first year of the COVID-19 pandemic (April 2020-March 2021) among patients with chronic conditions insured by Medicaid or Medicare Advantage. Methods: A prospective cohort study of 10,452 Kaiser Permanente Northern California members insured by Medicaid and 52,890 Kaiser Permanente Colorado members insured by Medicare Advantage was conducted. Difference-in-differences (DID) between the pre-COVID and COVID years in telehealth versus in-person health care utilization and adherence to chronic disease medicines by food insecurity and by physical limitation status were measured. Results: Food insecurity and physical limitations were each associated with small but significantly greater shifts from in-person to telehealth. Medicare Advantage members with physical limitations also had significantly greater decline in adherence to chronic medications from year to year compared with those without physical limitations (DID from pre-COVID year to COVID year ranged from 0.7% to 3.6% greater decline by medication class, p < 0.01). Conclusions: Food insecurity and physical limitations did not present significant barriers to the transition to telehealth during the COVID pandemic. The greater decrease in medication adherence among older patients with physical limitations suggests that care systems must further address the needs of this high-risk population.

Distinct healthcare utilization profiles of high healthcare use tuberculosis survivors: A latent class analysis

BACKGROUND

Recent data have demonstrated that healthcare use after treatment for respiratory tuberculosis (TB) remains elevated in the years following treatment completion. However, it remains unclear which TB survivors are high healthcare users and whether any variation exists within this population. Thus, the primary objective of this study was to identify distinct profiles of high healthcare-use TB survivors to help inform post-treatment support and care.

METHODS

Using linked health administrative data from British Columbia, Canada, we identified foreign-born individuals who completed treatment for incident respiratory TB between 1990 and 2019. We defined high healthcare-use TB survivors as those in the top 10% of annual emergency department visits, hospital admissions, or general practitioner visits among the study population during the five-year period immediately following TB treatment completion. We then used latent class analysis to categorize the identified high healthcare-use TB survivors into subgroups.

RESULTS

Of the 1,240 people who completed treatment for respiratory TB, 258 (20.8%) people were identified as high post- TB healthcare users. Latent class analysis results in a 2-class solution. Class 1 (n = 196; 76.0%) included older individuals (median age 71.0; IQR 59.8, 79.0) with a higher probability of pre-existing hypertension and diabetes (41.3% and 33.2%, respectively). Class 2 (n = 62; 24.0%) comprised of younger individuals (median age 31.0; IQR 27.0, 41.0) with a high probability (61.3%) of immigrating to Canada within five years of their TB diagnosis and a low probability (11.3%) of moderate to high continuity of primary care.

DISCUSSION

Our findings suggest that foreign-born high healthcare-use TB survivors in a high-resource setting may be categorized into distinct profiles to help guide the development of person-centred care strategies targeting the long-term health impacts TB survivors face.

Cluster analysis driven by unsupervised latent feature learning of medications to identify novel pharmacophenotypes of critically ill patients

Unsupervised clustering of intensive care unit (ICU) medications may identify unique medication clusters (i.e., pharmacophenotypes) in critically ill adults. We performed an unsupervised analysis with Restricted Boltzmann Machine of 991 medications profiles of patients managed in the ICU to explore pharmacophenotypes that correlated with ICU complications (e.g., mechanical ventilation) and patient-centered outcomes (e.g., length of stay, mortality). Six unique pharmacophenotypes were observed, with unique medication profiles and clinically relevant differences in ICU complications and patient-centered outcomes. While pharmacophenotypes 2 and 4 had no statistically significant difference in ICU length of stay, duration of mechanical ventilation, or duration of vasopressor use, their mortality differed significantly (9.0% vs. 21.9%, p < 0.0001). Pharmacophenotype 4 had a mortality rate of 21.9%, compared with the rest of the pharmacophenotypes ranging from 2.5 to 9%. Phenotyping approaches have shown promise in classifying the heterogenous syndromes of critical illness to predict treatment response and guide clinical decision support systems but have never included comprehensive medication information. This first-ever machine learning approach revealed differences among empirically-derived subgroups of ICU patients that are not typically revealed by traditional classifiers. Identification of pharmacophenotypes may enable enhanced decision making to optimize treatment decisions.

Opioid prescribing patterns among medical practitioners in New South Wales, Australia

INTRODUCTION

Prescriber behaviour is important for understanding opioid use patterns. We described variations in practitioner-level opioid prescribing in New South Wales, Australia (2013-2018).

METHODS

We quantified opioid prescribing patterns among medical practitioners using population-level dispensing claims data, and used partitioning around medoids to identify clusters of practitioners who prescribe opioids based on prescribing patterns and patient characteristics identified from linked dispensing claims, hospitalisations and mortality data.

RESULTS

The number of opioid prescribers ranged from 20,179 in 2013 to 23,408 in 2018. The top 1% of practitioners prescribed 15% of all oral morphine equivalent (OME) milligrams dispensed annually, with a median of 1382 OME grams (interquartile range [IQR], 1234-1654) per practitioner; the bottom 50% prescribed 1% of OMEs dispensed, with a median of 0.9 OME grams (IQR 0.2-2.6). Based on 63.6% of practitioners with ≥10 patients filling opioid prescriptions in 2018, we identified four distinct practitioner clusters. The largest cluster prescribed multiple analgesic medicines for older patients (23.7% of practitioners) accounted for 76.7% of all OMEs dispensed and comprised 93.0% of the top 1% of practitioners by opioid volume dispensed. The cluster prescribing analgesics for younger patients with high rates of surgery (18.7% of practitioners) prescribed only 1.6% of OMEs. The remaining two clusters comprised 21.2% of prescribers and 20.9% of OMEs dispensed.

DISCUSSION AND CONCLUSION

We observed substantial variation in opioid prescribing among practitioners, clustered around four general patterns. We did not assess appropriateness but some prescribing patterns are concerning. Our findings provide insights for targeted interventions to curb potentially harmful practices.

Identifying clusters of coexisting conditions and outcomes among adults admitted to hospital with community-acquired pneumonia: a multicentre cohort study

BACKGROUND

Little is known about patterns of coexisting conditions and their influence on clinical care or outcomes in adults admitted to hospital for community-acquired pneumonia (CAP). We sought to evaluate how coexisting conditions cluster in this population to advance understanding of how multimorbidity affects CAP.

METHODS

We studied 11 085 adults admitted to hospital with CAP at 7 hospitals in Ontario, Canada. Using cluster analysis, we identified patient subgroups based on clustering of comorbidities in the Charlson Comorbidity Index. We derived and replicated cluster analyses in independent cohorts (derivation sample 2010-2015, replication sample 2015-2017), then combined these into a total cohort for final cluster analyses. We described differences in medications, imaging and outcomes.

RESULTS

Patients clustered into 7 subgroups. The low comorbidity subgroup (n = 3052, 27.5%) had no comorbidities. The DM-HF-Pulm subgroup had prevalent diabetes, heart failure and chronic lung disease (n = 1710, 15.4%). One disease category defined each remaining subgroup, as follows: pulmonary (n = 1621, 14.6%), diabetes (n = 1281, 11.6%), heart failure (n = 1370, 12.4%), dementia (n = 1038, 9.4%) and cancer (n = 1013, 9.1%). Corticosteroid use ranged from 11.5% to 64.9% in the dementia and pulmonary subgroups, respectively. Piperacillin-tazobactam use ranged from 9.1% to 28.0% in the pulmonary and cancer subgroups, respectively. The use of thoracic computed tomography ranged from 5.7% to 36.3% in the dementia and cancer subgroups, respectively. Adjusting for patient factors, the risk of in-hospital death was greater in the cancer (adjusted odds ratio [OR] 3.12, 95% confidence interval [CI] 2.44-3.99), dementia (adjusted OR 1.57, 95% CI 1.05-2.35), heart failure (adjusted OR 1.66, 95% CI 1.35-2.03) and DM-HF-Pulm subgroups (adjusted OR 1.35, 95% CI 1.12-1.61), and lower in the diabetes subgroup (adjusted OR 0.67, 95% CI 0.50-0.89), compared with the low comorbidity group.

INTERPRETATION

Patients admitted to hospital with CAP cluster into clinically recognizable subgroups based on coexisting conditions. Clinical care and outcomes vary among these subgroups with little evidence to guide decision-making, highlighting opportunities for research to personalize care.

Variation in Outpatient Postpartum Care Use in the United States: A Latent Class Analysis

INTRODUCTION

Despite efforts to improve postpartum health care in the United States, little is known about patterns of postpartum care beyond routine postpartum visit attendance. This study aimed to describe variation in outpatient postpartum care patterns.

METHODS

In this longitudinal cohort study of national commercial claims data, we used latent class analysis to identify subgroups of patients (classes) with similar outpatient postpartum care patterns (defined by the number of preventive, problem, and emergency department outpatient visits in the 60 days after birth). We also compared classes in terms of maternal sociodemographics and clinical characteristics measured at childbirth, as well as total health spending and rates of adverse events (all-cause hospitalizations and severe maternal morbidity) measured from childbirth to the late postpartum period (61-365 days after birth).

RESULTS

The study cohort included 250,048 patients hospitalized for childbirth in 2016. We identified six classes with distinct outpatient postpartum care patterns in the 60 days after birth, which we classified into three broad groups: no care (class 1 [32.4% of the total sample]); preventive care only (class 2 [18.3%]); and problem care (classes 3-6 [49.3%]). The prevalence of clinical risk factors at childbirth increased progressively from class 1 to class 6; for example, 6.7% of class 1 patients had any chronic disease compared with 15.5% of class 5 patients. Severe maternal morbidity was highest among the high problem care classes (classes 5 and 6): 1.5% of class 6 patients experienced severe maternal morbidity in the postpartum period and 0.5% in the late postpartum period, compared with less than 0.1% of patients in classes 1 and 2.

CONCLUSIONS

Efforts to redesign and measure postpartum care should reflect the current heterogeneity in care patterns and clinical risks in the postpartum population.

Understanding categories of postpartum care use among privately insured patients in the United States: a cluster-analytic approach

The postpartum period is critical for the health and well-being of birthing people, yet little is known about the range of health care services and supports needed during this time. Maternity care patients are often targeted for clinical interventions based on "low risk" or "high risk" designations, but dichotomized measures can be imprecise and may not reflect meaningful groups for understanding needed postpartum care. Using claims data from privately insured patients with childbirths between 2016 and 2018, this study identifies categories and predictors of postpartum care utilization, including the use of maternal care and other, nonmaternal, care (eg, respiratory, digestive). We then compare identified utilization-based categories with typical high- and low-risk designations. Among 269 992 patients, 5 categories were identified: (1) low use (55% of births); (2) moderate maternal care use, low other care use (25%); (3) moderate maternal, high other (8%); (4) high maternal, moderate other (7%); and (5) high maternal, high other (5%). Utilization-based categories were better at differentiating postpartum care use and were more consistent across patient profiles, compared with high- and low-risk dichotomies. Identifying categories of postpartum care need beyond a simple risk dichotomy is warranted and can assist in maternal health services research, policymaking, and clinical practice.

Predicting outcomes at the individual patient level: what is the best method?

OBJECTIVE

When developing prediction models, researchers commonly employ a single model which uses all the available data (end-to-end approach). Alternatively, a similarity-based approach has been previously proposed, in which patients with similar clinical characteristics are first grouped into clusters, then prediction models are developed within each cluster. The potential advantage of the similarity-based approach is that it may better address heterogeneity in patient characteristics. However, it remains unclear whether it improves the overall predictive performance. We illustrate the similarity-based approach using data from people with depression and empirically compare its performance with the end-to-end approach.

METHODS

We used primary care data collected in general practices in the UK. Using 31 predefined baseline variables, we aimed to predict the severity of depressive symptoms, measured by Patient Health Questionnaire-9, 60 days after initiation of antidepressant treatment. Following the similarity-based approach, we used k-means to cluster patients based on their baseline characteristics. We derived the optimal number of clusters using the Silhouette coefficient. We used ridge regression to build prediction models in both approaches. To compare the models' performance, we calculated the mean absolute error (MAE) and the coefficient of determination (R2) using bootstrapping.

RESULTS

We analysed data from 16 384 patients. The end-to-end approach resulted in an MAE of 4.64 and R2 of 0.20. The best-performing similarity-based model was for four clusters, with MAE of 4.65 and R2 of 0.19.

CONCLUSIONS

The end-to-end and the similarity-based model yielded comparable performance. Due to its simplicity, the end-to-end approach can be favoured when using demographic and clinical data to build prediction models on pharmacological treatments for depression.

Differential Effects of an Emergency Department-to-Home Care Transitions Intervention in an Older Adult Population: A Latent Class Analysis

BACKGROUND

Older adults frequently return to the emergency department (ED) within 30 days of a visit. High-risk patients can differentially benefit from transitional care interventions. Latent class analysis (LCA) is a model-based method used to segment the population and test intervention effects by subgroup.

OBJECTIVES

We aimed to identify latent classes within an older adult population from a randomized controlled trial evaluating the effectiveness of an ED-to-home transitional care program and test whether class membership modified the intervention effect.

RESEARCH DESIGN

Participants were randomized to receive the Care Transitions Intervention or usual care. Study staff collected outcomes data through medical record reviews and surveys. We performed LCA and logistic regression to evaluate the differential effects of the intervention by class membership.

SUBJECTS

Participants were ED patients (age 60 y and above) discharged to a community residence.

MEASURES

Indicator variables for the LCA included clinically available and patient-reported data from the initial ED visit. Our primary outcome was ED revisits within 30 days. Secondary outcomes included ED revisits within 14 days, outpatient follow-up within 7 and 30 days, and self-management behaviors.

RESULTS

We interpreted 6 latent classes in this study population. Classes 1, 4, 5, and 6 showed a reduction in ED revisit rates with the intervention; classes 2 and 3 showed an increase in ED revisit rates. In class 5, we found evidence that the intervention increased outpatient follow-up within 7 and 30 days (odds ratio: 1.81, 95% CI: 1.13-2.91; odds ratio: 2.24, 95% CI: 1.25-4.03).

CONCLUSIONS

Class membership modified the intervention effect. Population segmentation is an important step in evaluating a transitional care intervention.

Population Segmentation for COVID-19 Vaccine Outreach: A Clustering Analysis and Implementation in Missouri

OBJECTIVES

The purpose of this work was to segment the Missouri population into unique groups related to COVID-19 vaccine acceptance using data science and behavioral science methods to develop tailored vaccine outreach strategies.

METHODS

Cluster analysis techniques were applied to a large data set that aggregated vaccination data with behavioral and demographic data from the American Community Survey and Deloitte's HealthPrism™ data set. Outreach recommendations were developed for each cluster, specific to each group's practical and motivational barriers to vaccination.

RESULTS

Following selection procedures, 10 clusters-or segments-of census tracts across Missouri were identified on the basis of k-means clustering analysis of 18 different variables. Each cluster exhibited unique geographic, demographic, socioeconomic, and behavioral patterns, and outreach strategies were developed on the basis of each cluster's practical and motivational barriers.

DISCUSSION

The segmentation analysis served as the foundation for "working groups" comprising the 115 local public health agencies (LPHAs) across the state. LPHAs with similar community segments in their service area were grouped together to discuss their communities' specific challenges, share lessons learned, and brainstorm new approaches. The working groups provided a novel way for public health to organize and collaborate across the state. Widening the aperture beyond Missouri, population segmentation via cluster analysis is a promising approach for public health practitioners interested in developing a richer understanding of the types of populations they serve. By pairing segmentation with behavioral science, practitioners can develop outreach programs and communications campaigns that are personalized to the specific behavioral barriers and needs of the population in focus. While our work focused on COVID-19, this approach has broad applicability to enhance the way public health practitioners understand the populations they serve to deliver more tailored services.

Novel Metabolic Subtypes in Pregnant Women and Risk of Early Childhood Obesity in Offspring

Importance

The in utero metabolic milieu is associated with offspring adiposity. Standard definitions of maternal obesity (according to prepregnancy body mass index [BMI]) and gestational diabetes (GDM) may not be adequate to capture subtle yet important differences in the intrauterine environment that could be involved in programming.

Objectives

To identify maternal metabolic subgroups during pregnancy and to examine associations of subgroup classification with adiposity traits in their children.

Design, Setting, and Participants

This cohort study included mother-offspring pairs in the Healthy Start prebirth cohort (enrollment: 2010-2014) recruited from University of Colorado Hospital obstetrics clinics in Aurora, Colorado. Follow-up of women and children is ongoing. Data were analyzed from March to December 2022.

Exposures

Metabolic subtypes of pregnant women ascertained by applying k-means clustering on 7 biomarkers and 2 biomarker indices measured at approximately 17 gestational weeks: glucose, insulin, Homeostatic Model Assessment for Insulin Resistance, total cholesterol, high-density lipoprotein cholesterol (HDL-C), triglycerides, free fatty acids (FFA), HDL-C:triglycerides ratio, and tumor necrosis factor α.

Main Outcomes and Measures

Offspring birthweight z score and neonatal fat mass percentage (FM%). In childhood at approximately 5 years of age, offspring BMI percentile, FM%, BMI in the 95th percentile or higher, and FM% in the 95th percentile or higher.

Results

A total of 1325 pregnant women (mean [SD] age, 27.8 [6.2 years]; 322 [24.3%] Hispanic, 207 non-Hispanic Black [15.6%], and 713 [53.8%] non-Hispanic White), and 727 offspring with anthropometric data measured in childhood (mean [SD] age 4.81 [0.72] years, 48% female) were included. We identified the following 5 maternal metabolic subgroups: reference (438 participants), high HDL-C (355 participants), dyslipidemic-high triglycerides (182 participants), dyslipidemic-high FFA (234 participants), and insulin resistant (IR)-hyperglycemic (116 participants). Compared with the reference subgroup, women in the IR-hyperglycemic and dyslipidemic-high FFA subgroups had offspring with 4.27% (95% CI, 1.94-6.59) and 1.96% (95% CI, 0.45-3.47) greater FM% during childhood, respectively. There was a higher risk of high FM% among offspring of the IR-hyperglycemic (relative risk, 8.7; 95% CI, 2.7-27.8) and dyslipidemic-high FFA (relative risk, 3.4; 95% CI, 1.0-11.3) subgroups; this risk was of greater magnitude compared with prepregnancy obesity alone, GDM alone, or both conditions.

Conclusions and Relevance

In this cohort study, an unsupervised clustering approach revealed distinct metabolic subgroups of pregnant women. These subgroups exhibited differences in risk of offspring adiposity in early childhood. Such approaches have the potential to refine understanding of the in utero metabolic milieu, with utility for capturing variation in sociocultural, anthropometric, and biochemical risk factors for offspring adiposity.

Three distinct symptom profiles among older adults with severe dementia: A latent class analysis

OBJECTIVES

Older adults with severe dementia experience multiple symptoms at the end of life. This study aimed to delineate distinct symptom profiles of older adults with severe dementia and to assess their association with older adults' and caregiver characteristics and 1-year mortality among older adults.

METHODS

We used baseline data from a cohort of 215 primary informal caregivers of older adults with severe dementia in Singapore. We identified 10 indicators representing physical, emotional, and functional symptoms, and responsive behaviors, and conducted latent class analysis. We assessed the association between delineated older adults' symptom profiles and their use of potentially burdensome health-care interventions in the past 4 months; older adults' 1-year mortality; and caregiver outcomes.

RESULTS

We delineated 3 profiles of older adults - primarily responsive behaviors (Class 1; 33%); physical and emotional symptoms with responsive behaviors (Class 2; 20%); and high functional deficits with loss of speech and eye contact (Class 3; 47%). Classes 2 and 3 older adults were more likely to have received a potentially burdensome intervention for symptoms in the past 4 months and have a greater hazard for 1-year mortality. Compared to Class 1, caregivers of Class 2 older adults were more likely to experience adverse caregiver outcomes, that is, higher distress, impact on schedule and health, anticipatory grief, and coping and lower satisfaction with care received (p<0.01 for all).

SIGNIFICANCE OF RESULTS

The 3 delineated profiles of older adults can be used to plan or optimize care plans to effectively manage symptoms of older adults and improve their caregivers' outcomes.

Phenomapping in heart failure with preserved ejection fraction: insights, limitations, and future directions

Heart failure with preserved ejection fraction (HFpEF) is a heterogeneous entity with complex pathophysiology and manifestations. Phenomapping is the process of applying statistical learning techniques to patient data to identify distinct subgroups based on patterns in the data. Phenomapping has emerged as a technique with potential to improve the understanding of different HFpEF phenotypes. Phenomapping efforts have been increasing in HFpEF over the past several years using a variety of data sources, clinical variables, and statistical techniques. This review summarizes methodologies and key takeaways from these studies, including consistent discriminating factors and conserved HFpEF phenotypes. We argue that phenomapping results to date have had limited implications for clinical care and clinical trials, given that the phenotypes, as currently described, are not reliably identified in each study population and may have significant overlap. We review the inherent limitations of aggregating and utilizing phenomapping results. Lastly, we discuss potential future directions, including using phenomapping to optimize the likelihood of clinical trial success or to drive discovery in mechanisms of the disease process of HFpEF.

Outcomes of home-based primary care for homebound older adults: A randomized clinical trial

BACKGROUND

Homebound older adults are medically complex and often have difficulty accessing outpatient medical care. Home-based primary care (HBPC) may improve care and outcomes for this population but data from randomized trials of HBPC in the United States are limited.

METHODS

We conducted a randomized controlled trial of HBPC versus office-based primary care for adults ages ≥65 years who reported ≥1 hospitalization in the prior 12 months and met the Medicare definition of homebound. HBPC was provided by teams consisting of a physician, nurse practitioner, nurse, and social worker. Data were collected at baseline, 6- and 12-months. Outcomes were quality of life, symptoms, satisfaction with care, hospitalizations, and emergency department (ED) visits. Recruitment was terminated early because more deaths were observed for intervention patients.

RESULTS

The study enrolled 229 patients, 65.4% of planned recruitment. The mean age was 82 (9.0) years and 72.3% had dementia. Of those assigned to HBPC, 34.2% never received it. Intervention patients had greater satisfaction with care than controls (2.26, 95% CI 1.46-3.06, p < 0.0001; effect size 0.74) and lower hospitalization rates (-17.9%, 95% CI -31.0% to -1.0%; p = 0.001; number needed to treat 6, 95% CI 3-100). There were no significant differences in quality of life (1.25, 95% CI -0.39-2.89, p = 0.13), symptom burden (-1.92, 95% CI -5.22-1.37, p = 0.25) or ED visits (1.2%, 95% CI -10.5%-12.4%; p = 0.87). There were 24 (21.1%) deaths among intervention patients and 12 (10.7%) among controls (p < 0.0001).

CONCLUSION

HBPC was associated with greater satisfaction with care and lower hospitalization rates but also more deaths compared to office-based primary care. Additional research is needed to understand the nature of the higher death rate for HBPC patients, as well as to determine the effects of HBPC on quality of life and symptom burden given the trial's early termination.

Prevalence of use and characteristics of users of home-based medical care in Medicare Advantage

BACKGROUND/OBJECTIVES

Home-based medical care (HBMC) is longitudinal medical care provided by physicians, advanced practice providers, and, often, inter-professional care teams to patients in their homes. Our objective is to determine the prevalence of HBMC among older adults (≥65) insured by a Medicare Advantage (MA) plan and compare characteristics of those who receive HBMC to those who do not.

METHODS

Study used de-identified medical claims and enrollment records for MA beneficiaries during calendar years 2017 and 2018 linked with socioeconomic status data in the OptumLabs Data Warehouse. We defined a cohort of MA beneficiaries age ≥65 receiving HBMC for at least 2 months during 2017-2018, described the cohort using demographic, utilization, and comorbidity data and compared it to a 5% random sample of a population of MA beneficiaries age ≥65 not receiving HBMC (No HBMC).

RESULTS

Overall, 1.45% of the study cohort age ≥65 received HBMC. Compared to No HBMC (n = 132,147), those receiving HBMC (n = 38,800) were more likely to be: older (46.6% vs. 11.9% age 85+); female (70.8% vs. 58.5%); Black (12.3% vs. 11.3%); urban (90.3% vs. 81.3%); experience hospitalization (38.0% vs. 13.3%), emergency department visit (58.3% vs. 26.9%), ambulance trip (44.1% vs. 9.6%), skilled nursing facility (37.6% vs. 6.4%), or hospice care admission (21.1% vs. 3.5%). They also were more likely to experience a wide range of chronic conditions including dementia (58.1% vs. 5.2%), morbidity burden (Charlson score 3.4 vs. 1.8), and serious illness (77.1% vs. 29.5%). All comparisons p < 0.0001.

CONCLUSIONS

MA beneficiaries who received HBMC are older, experience greater chronic and serious illness burden, and higher levels of facility-based care than those who did not receive HBMC. MA plans need strategies to identify patients that would benefit from HBMC and develop approaches to deliver such care to this impactful, often invisible population.

Defining clinical subtypes of adult asthma using electronic health records: Analysis of a large UK primary care database with external validation

INTRODUCTION

Asthma is one of the commonest chronic conditions in the world. Subtypes of asthma have been defined, typically from clinical datasets on small, well-characterised subpopulations of asthma patients. We sought to define asthma subtypes from large longitudinal primary care electronic health records (EHRs) using cluster analysis.

METHODS

In this retrospective cohort study, we extracted asthma subpopulations from the Optimum Patient Care Research Database (OPCRD) to robustly train and test algorithms, and externally validated findings in the Secure Anonymised Information Linkage (SAIL) Databank. In both databases, we identified adults with an asthma diagnosis code recorded in the three years prior to an index date. Train and test datasets were selected from OPCRD using an index date of Jan 1, 2016. Two internal validation datasets were selected from OPCRD using index dates of Jan 1, 2017 and 2018. Three external validation datasets were selected from SAIL using index dates of Jan 1, 2016, 2017 and 2018. Each dataset comprised 50,000 randomly selected non-overlapping patients. Subtypes were defined by applying multiple correspondence analysis and k-means cluster analysis to the train dataset, and were validated in the internal and external validation datasets.

RESULTS

We defined six asthma subtypes with clear clinical interpretability: low inhaled corticosteroid (ICS) use and low healthcare utilisation (30% of patients); low-to-medium ICS use (36%); low-to-medium ICS use and comorbidities (12%); varied ICS use and comorbid chronic obstructive pulmonary disease (4%); high (10%) and very high ICS use (7%). The subtypes were replicated with high accuracy in internal (91-92%) and external (84-86%) datasets.

CONCLUSION

Asthma subtypes derived and validated in large independent EHR databases were primarily defined by level of ICS use, level of healthcare use, and presence of comorbidities. This has important clinical implications towards defining asthma subtypes, facilitating patient stratification, and developing more personalised monitoring and treatment strategies.

Clinical Phenotypes of Immediate First-Dose Reactions to mRNA COVID-19: A Multicenter Latent Class Analysis

BACKGROUND

Although immediate potentially allergic reactions have been reported after dose 1 of mRNA coronavirus disease 2019 (COVID-19) vaccines, comprehensively defined subtypes have not been clearly distinguished.

OBJECTIVE

To define distinct clinical phenotypes of immediate reactions after dose 1 of mRNA COVID-19 vaccination, and to assess the relation of clinical phenotype to mRNA COVID-19 vaccine second dose tolerance.

METHODS

This retrospective study included patients with 1 or more potentially allergic symptoms or signs within 4 hours of receiving dose 1 of an mRNA COVID-19 vaccine and assessed by allergy/immunology specialists from 5 U.S. academic medical centers (January-June 2021). We used latent class analysis-an unbiased, machine-learning modeling method-to define novel clinical phenotypes. We assessed demographic, clinical, and reaction characteristics associated with phenotype membership. Using log-binomial regression, we assessed the relation between phenotype membership and second dose tolerance, defined as either no symptoms or mild, self-limited symptoms resolving with antihistamines alone. A sensitivity analysis considered second dose tolerance as objective signs only.

RESULTS

We identified 265 patients with dose-1 immediate reactions with 3 phenotype clusters: (1) Limited or Predominantly Cutaneous, (2) Sensory, and (3) Systemic. A total of 223 patients (84%) received a second dose and 200 (90%) tolerated their second dose. Sensory cluster (all patients had the symptom of numbness or tingling) was associated with a higher likelihood of second dose intolerance, but this finding did not persist when accounting for objective signs.

CONCLUSIONS

Three novel clinical phenotypes of immediate-onset reactions after dose 1 of mRNA COVID-19 vaccines were identified using latent class analysis: (1) Limited or Predominantly Cutaneous, (2) Sensory, and (3) Systemic. Whereas these clinical phenotypes may indicate differential mechanistic etiologies or associations with subsequent dose tolerance, most individuals proceeding to their second dose tolerated it.

Cut points of the conicity index as an indicator of abdominal obesity in individuals undergoing hemodialysis: An analysis of latent classes

OBJECTIVES

Abdominal obesity favors the involvement of cardiometabolic complications in renal patients on hemodialysis. Thus, the aim of the study was to identify the cut-points of the conicity index in individuals undergoing hemodialysis.

METHODS

This was a cross-sectional study carried out with 953 individuals undergoing hemodialysis in clinics in a metropolitan region of southeastern Brazil. The conicity index was calculated using the following mathematical equation: waist circumference/0.109 × √weight/height. The receiver operating characteristic (ROC) curve was calculated from the analysis of latent classes by cross-validation through a latent variable of abdominal obesity. This latent variable was defined using the response pattern of the observed anthropometric variables considering the presence and absence of abdominal obesity: waist circumference, waist-to-height ratio, and body shape index. The cut-points identified were elucidated by the area under the curve (AUC), Youden index, sensitivity, and specificity.

RESULTS

The cut-points for the conicity index found for both sexes were similar, resulting in a cut-point for men of 1.275 (AUC, 0.921; Youden index, 0.666), with a sensitivity and specificity of 83% and 83.6%, and a cut-point for women of 1.285 (AUC, 0.921; Youden index, 0.679), with a sensitivity and specificity of 78.6% and 89.3%, respectively.

CONCLUSIONS

The conicity index showed high discriminatory power for the identification of abdominal obesity in hemodialysis patients, therefore it can be a simple and easily accessible tool to be incorporated into clinical practice in this population.

Illness severity characteristics and outcomes of patients remaining on an acute ward following medical emergency team review: a latent profile analysis

BACKGROUND

Patients requiring medical emergency team (MET) review have complex clinical needs, and most remain on the ward after review. Current detection instruments cannot identify post-MET patient requirements, meaning patients remain undistinguished, potentially resulting in missed management opportunities. We propose that deteriorating patients will cluster along dimensions of illness severity and that these clusters may be used to strengthen patient risk management practices.

OBJECTIVE

To identify and define the number of illness severity clusters and report outcomes among ward patients following MET review.

STUDY DESIGN AND SETTING

This retrospective cohort study examined the clinical records of 1500 adult ward patients following MET review at an Australian quaternary hospital. Three-step latent profile analysis methods were used to determine clusters using Sequential Organ Failure Assessment (SOFA) and Nursing Activities Score (NAS) as illness severity indicators. Study outcomes were (1) hospital mortality, (2) unplanned intensive care unit (ICU) admission and (3) subsequent MET review.

RESULTS

Patients were unplanned (73.9%) and medical (57.5%) admissions with at least one comorbidity (51.4%), and complex combinations of acuity (SOFA range 1-17) and dependency (NAS range 22.4%-148.5%). Five clusters are reported. Patients in cluster 1 were equivalent to clinically stable general ward patients. Organ failure and complexity increased with cluster progression-clusters 2 and 3 were equivalent to subspecialty/higher-dependency wards, and clusters 4 and 5 were equivalent to ICUs. Patients in cluster 5 had the greatest odds for death (OR 26.2, 95% CI 23.3 to 31.3), unplanned ICU admission (OR 3.1, 95% CI 3.0 to 3.1) and subsequent MET review (OR 2.4, 95% CI 2.4 to 2.6).

CONCLUSION

The five illness severity clusters may be used to define patients at risk of poorer outcomes who may benefit from enhanced levels of monitoring and targeted care.

Complex early childhood experiences: Characteristics of Northern Territory children across health, education and child protection data

Early identification of vulnerable children to protect them from harm and support them in achieving their long-term potential is a community priority. This is particularly important in the Northern Territory (NT) of Australia, where Aboriginal children are about 40% of all children, and for whom the trauma and disadvantage experienced by Aboriginal Australians has ongoing intergenerational impacts. Given that shared social determinants influence child outcomes across the domains of health, education and welfare, there is growing interest in collaborative interventions that simultaneously respond to outcomes in all domains. There is increasing recognition that many children receive services from multiple NT government agencies, however there is limited understanding of the pattern and scale of overlap of these services. In this paper, NT health, education, child protection and perinatal datasets have been linked for the first time. The records of 8,267 children born in the NT in 2006-2009 were analysed using a person-centred analytic approach. Unsupervised machine learning techniques were used to discover clusters of NT children who experience different patterns of risk. Modelling revealed four or five distinct clusters including a cluster of children who are predominantly ill and experience some neglect, a cluster who predominantly experience abuse and a cluster who predominantly experience neglect. These three, high risk clusters all have low school attendance and together comprise 10-15% of the population. There is a large group of thriving children, with low health needs, high school attendance and low CPS contact. Finally, an unexpected cluster is a modestly sized group of non-attendees, mostly Aboriginal children, who have low school attendance but are otherwise thriving. The high risk groups experience vulnerability in all three domains of health, education and child protection, supporting the need for a flexible, rather than strictly differentiated response. Interagency cooperation would be valuable to provide a suitably collective and coordinated response for the most vulnerable children.

The role of trauma and positive youth development in polysubstance use among rural middle school students: a latent class analysis

BACKGROUND

Rural youth often begin developing polysubstance use and other risk behaviors during middle school. However, little polysubstance use research focuses on rural middle school youth. Our research uses Latent Class Analysis to understand existing patterns of rural middle school polysubstance use and risk and protective factors associated with polysubstance use.

METHODS

We used survey data from a rural middle school pregnancy prevention program (N = 2,708). The survey included measures of demographics, lifetime substance use, trauma (adverse childhood experiences and bullying victimization) and aspects of youth development (parent communication on drugs and alcohol, parent connectedness and school connectedness). We used latent class analysis to produce participant polysubstance use profiles and multinomial regression to examine associations between polysubstance use, demographics, trauma and aspects of youth development.

RESULTS

We categorized our participants into four latent classes. Our analysis classified 2.2% of participants as Regular Polysubstance users, 6.9% as Polysubstance experimenters, 19% as Vape + Alcohol experimenters and 71.9% as Non-Users. More adverse childhood experiences were associated with greater risk of polysubstance use and experimentation. Bullying was positively associated with greater risk of vape and alcohol experimentation. Higher reported parental and school connectedness were associated with lower risk of high frequency polysubstance use. Higher reported school connection was also associated with lower risk of polysubstance experimentation.

CONCLUSION

Rural substance use prevention programs should begin during middle school, as polysubstance use development is common among rural middle schoolers. These programs should be trauma informed and focus on connectedness as a modifiable factor to reduce risk of polysubstance use development.

TRIAL REGISTRATION

This article does not report results of a health care intervention on human participants.

Characteristics of High-Need, High-Cost Patients : A "Best-Fit" Framework Synthesis

BACKGROUND

Accurately identifying high-need, high-cost (HNHC) patients to reduce their preventable or modifiable health care use for their chronic conditions is a priority and a challenge for U.S. policymakers, health care delivery systems, and payers.

PURPOSE

To identify characteristics and criteria to distinguish HNHC patients.

DATA SOURCES

Searches of multiple databases and gray literature from 1 January 2000 to 22 January 2022.

STUDY SELECTION

English-language studies of characteristics and criteria to identify HNHC adult patients, defined as those with high use (emergency department, inpatient, or total services) or high cost.

DATA EXTRACTION

Independent, dual-review extraction and quality assessment.

DATA SYNTHESIS

The review included 64 studies comprising multivariate exposure studies (n = 47), cluster analyses (n = 11), and qualitative studies (n = 6). A National Academy of Medicine (NAM) taxonomy was an initial "best-fit" framework for organizing the synthesis of the findings. Patient characteristics associated with being HNHC included number and severity of comorbid conditions and having chronic clinical conditions, particularly heart disease, chronic kidney disease, chronic lung disease, diabetes, cancer, and hypertension. Patients' risk for being HNHC was often amplified by behavioral health conditions and social risk factors. The reviewers revised the NAM taxonomy to create a final framework, adding chronic pain and prior patterns of high health care use as characteristics associated with an increased risk for being HNHC.

LIMITATION

Little evidence distinguished potentially preventable or modifiable health care use from overall use.

CONCLUSION

A combination of characteristics can be useful for identifying HNHC patients. Because of the complexity of their conditions and circumstances, improving their quality of care will likely also require an individualized assessment of care needs and availability of support services.

PRIMARY FUNDING SOURCE

Agency for Healthcare Research and Quality. (PROSPERO: CRD42020161179).

Cluster Analysis of the Highest Users of Medical, Behavioral Health, and Social Services in San Francisco

BACKGROUND

In the City and County of San Francisco, frequent users of emergent and urgent services across different settings (i.e., medical, mental health (MH), substance use disorder (SUD) services) are referred to as high users of multiple systems (HUMS). While often grouped together, frequent users of the health care system are likely a heterogenous population composed of subgroups with differential management needs.

OBJECTIVE

To identify subgroups within this HUMS population using a cluster analysis.

DESIGN

Cross-sectional study of HUMS patients for the 2019-2020 fiscal year using the Coordinated Care Management System (CCMS), San Francisco Department of Public Health's integrated data system.

PARTICIPANTS

We calculated use scores based on nine types of urgent and emergent medical, MH, and SUD services and identified the top 5% of HUMS patients. Through k-medoids cluster analysis, we identified subgroups of HUMS patients.

MAIN MEASURES

Subgroup-specific demographic, comorbidity, and service use profiles.

KEY RESULTS

The top 5% of HUMS patients in the study period included 2657 individuals; 69.7% identified as men and 66.5% identified as non-White. We detected 5 subgroups: subgroup 1 (N = 298, 11.2%) who were relatively younger with prevalent MH and SUD comorbidities, and MH services use; subgroup 2 (N = 478, 18.0%), who were experiencing homelessness, with multiple comorbidities, and frequent use of medical services; subgroup 3 (N = 449, 16.9%), who disproportionately self-identified as Black, with prolonged homelessness, multiple comorbidities, and persistent HUMS status; subgroup 4 (N = 690, 26.0%), who were relatively older, disproportionately self-identified as Black, with prior homelessness, multiple comorbidities, and frequent use of medical services; and subgroup 5 (N=742, 27.9%), who disproportionately self-identified as Latinx, were housed, with medical comorbidities and frequent medical service use.

CONCLUSIONS

Our study highlights the heterogeneity of HUMS patients. Interventions must be tailored to meet the needs of these diverse patient subgroups.

Two-year change in latent classes of comorbidity among high-risk Veterans in primary care: a brief report

Background

Segmentation models such as latent class analysis are an increasingly popular approach to inform group-tailored interventions for high-risk complex patients. Multiple studies have identified clinically meaningful high-risk segments, but few have evaluated change in groupings over time.

Objectives

To describe population-level and individual change over time in latent comorbidity groups among Veterans at high-risk of hospitalization in the Veterans Health Administration (VA).

Research design

Using a repeated cross-sectional design, we conducted a latent class analysis of chronic condition diagnoses. We compared latent class composition, patient high-risk status, and patient class assignment in 2018 to 2020.

Subjects

Two cohorts of eligible patients were selected: those active in VA primary care and in the top decile of predicted one-year hospitalization risk in 2018 (n = 951,771) or 2020 (n = 978,771).

Measures

Medical record data were observed from January 2016–December 2020. Latent classes were modeled using indicators for 26 chronic health conditions measured with a 2-year lookback period from study entry.

Results

Five groups were identified in both years, labeled based on high prevalence conditions: Cardiometabolic (23% in 2018), Mental Health (18%), Substance Use Disorders (16%), Low Diagnosis (25%), and High Complexity (10%). The remaining 8% of 2018 patients were not assigned to a group due to low predicted probability. Condition prevalence overall and within groups was stable between years. However, among the 563,725 patients identified as high risk in both years, 40.8% (n = 230,185) had a different group assignment in 2018 versus 2020.

Conclusions

In a repeated latent class analysis of nearly 1 million Veterans at high-risk for hospitalization, population-level groups were stable over two years, but individuals often moved between groups. Interventions tailored to latent groups need to account for change in patient status and group assignment over time.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12913-022-08757-x.

Phenotype clustering in health care: A narrative review for clinicians

Human pathophysiology is occasionally too complex for unaided hypothetical-deductive reasoning and the isolated application of additive or linear statistical methods. Clustering algorithms use input data patterns and distributions to form groups of similar patients or diseases that share distinct properties. Although clinicians frequently perform tasks that may be enhanced by clustering, few receive formal training and clinician-centered literature in clustering is sparse. To add value to clinical care and research, optimal clustering practices require a thorough understanding of how to process and optimize data, select features, weigh strengths and weaknesses of different clustering methods, select the optimal clustering method, and apply clustering methods to solve problems. These concepts and our suggestions for implementing them are described in this narrative review of published literature. All clustering methods share the weakness of finding potential clusters even when natural clusters do not exist, underscoring the importance of applying data-driven techniques as well as clinical and statistical expertise to clustering analyses. When applied properly, patient and disease phenotype clustering can reveal obscured associations that can help clinicians understand disease pathophysiology, predict treatment response, and identify patients for clinical trial enrollment.

Clinical Outcome and Utilization Profiles Among Latent Groups of High-Risk Patients: Moving from Segmentation Towards Intervention

BACKGROUND

The ability of latent class models to identify clinically distinct groups among high-risk patients has been demonstrated, but it is unclear how healthcare data can inform group-specific intervention design.

OBJECTIVE

Examine how utilization patterns across latent groups of high-risk patients provide actionable information to guide group-specific intervention design.

DESIGN

Cohort study using data from 2012 to 2015.

PATIENTS

Participants were 934,787 patients receiving primary care in the Veterans Health Administration, with predicted probability of 12-month hospitalization in the top 10^th percentile during 2014.

MAIN MEASURES

Patients were assigned to latent groups via mixture-item response theory models based on 28 chronic conditions. We modeled odds of all-cause mortality, hospitalizations, and 30-day re-hospitalizations by group membership. Detailed outpatient and inpatient utilization patterns were compared between groups.

KEY RESULTS

A total of 764,257 (81.8%) of patients were matched with a comorbidity group. Groups were characterized by substance use disorders (14.0% of patients assigned), cardiometabolic conditions (25.7%), mental health conditions (17.6%), pain/arthritis (19.1%), cancer (15.3%), and liver disease (8.3%). One-year mortality ranged from 2.7% in the Mental Health group to 14.9% in the Cancer group, compared to 8.5% overall. In adjusted models, group assignment predicted significantly different odds of each outcome. Groups differed in their utilization of multiple types of care. For example, patients in the Pain group had the highest utilization of in-person primary care, with a mean (SD) of 5.3 (5.0) visits in the year of follow-up, while the Substance Use Disorder group had the lowest, with 3.9 (4.1) visits. The Substance Use Disorder group also had the highest rates of using services for housing instability (25.1%), followed by the Liver group (10.1%).

CONCLUSIONS

Latent groups of high-risk patients had distinct hospitalization and utilization profiles, despite having comparable levels of predicted baseline risk. Utilization profiles pointed towards system-specific care needs that could inform tailored interventions.

Exploring clinically relevant risk profiles in patients undergoing lumbar spinal fusion: a cohort study

Purpose

To explore risk profiles of patients scheduled for lumbar spinal fusion (LSF) and their association with short-term recovery of patient after surgery.

Methods

Forty-nine patients scheduled for elective 1–3 level LSF between March 2019 and June 2020 were included. Patients underwent a preoperative risk screening, consisting of an anamnesis, questionnaires and physical performance tests. A latent profile analysis (LPA) was used to identify possible risk profiles within this population.

Results

Two risk profiles could be established: a fit and deconditioned risk profile. A significant between-profile difference was found in smoking status (p = 0.007), RAND36-PCS (p < 0.001), Timed Up and Go (TUG) (p < 0.001), de Morton Morbidity Index (DEMMI) (p < 0.001), finger floor distance (p = 0.050), motor control (p = 0.020) and steep ramp test (p = 0.005). Moreover, the fit risk profile had a significant shorter time to functional recovery (3.65 days versus 4.89 days, p = 0.013) and length of hospital stay (5.06 days versus 6.00 days, p = 0.008) compared to the deconditioned risk profile. No differences in complication rates between both risk profiles could be established. Allocation to a risk profile was associated with the functional recovery rate (p = 0.042), but not with LOS or complications.

Conclusion

This study found a fit and deconditioned risk profile. The patients with a fit risk profile perceived a better quality of life, performed better in mobility, motor control, cardiopulmonary tests and showed also a significant shorter stay in the hospital and a shorter time to functional recovery. Preoperatively establishing a patient’s risk profile could aid in perioperative care planning and preoperative decision-making.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00586-022-07325-5.

The Identification of Subphenotypes and Associations with Health Outcomes in Patients with Opioid-Related Emergency Department Encounters Using Latent Class Analysis

The emergency department (ED) is a critical setting for the treatment of patients with opioid misuse. Detecting relevant clinical profiles allows for tailored treatment approaches. We sought to identify and characterize subphenotypes of ED patients with opioid-related encounters. A latent class analysis was conducted using 14,057,302 opioid-related encounters from 2016 through 2017 using the National Emergency Department Sample (NEDS), the largest all-payer ED database in the United States. The optimal model was determined by face validity and information criteria-based metrics. A three-step approach assessed class structure, assigned individuals to classes, and examined characteristics between classes. Class associations were determined for hospitalization, in-hospital death, and ED charges. The final five-class model consisted of the following subphenotypes: Chronic pain (class 1); Alcohol use (class 2); Depression and pain (class 3); Psychosis, liver disease, and polysubstance use (class 4); and Pregnancy (class 5). Using class 1 as the reference, the greatest odds for hospitalization occurred in classes 3 and 4 (Ors 5.24 and 5.33, p < 0.001) and for in-hospital death in class 4 (OR 3.44, p < 0.001). Median ED charges ranged from USD 2177 (class 1) to USD 2881 (class 4). These subphenotypes provide a basis for examining patient-tailored approaches for this patient population.

Exposure to Per- and Polyfluoroalkyl Substances and Mortality in U.S. Adults: A Population-Based Cohort Study

BACKGROUND

Per- and polyfluoroalkyl substances (PFAS) are widespread environmental contaminants associated with diseases such as cancer and dyslipidemia. However, few studies have investigated the association between PFAS mixture exposure and mortality in general populations.

OBJECTIVES

This study aimed to explore the association between PFAS mixture, perfluorooctanoic acid (PFOA), and perfluorooctane sulfonic acid (PFOS) and mortality in U.S. adults by a nationally representative cohort.

METHODS

Adults ≥18 years of age who were enrolled in the National Health and Nutrition Examination Survey (NHANES) (1999-2014) were included in our study. Baseline serum concentrations of seven PFAS were measured and individuals were followed up to 31 December 2015. Hazard ratios (HRs) and confidence intervals (CIs) were estimated using Cox proportional hazards models. Association between PFAS mixture exposure and mortality was analyzed using the k-means method by clustering PFAS mixtures into subgroups. Association between PFOA/PFOS exposure and mortality was subsequently analyzed in both continuous and categorical models.

RESULTS

During the follow-up period, 1,251 participants died. In the mixture analysis, the k-means algorithm clustered participants into low-, medium-, and high-exposure groups. Compared with the low-exposure group, participants in the high-exposure group showed significantly higher risks for all-cause mortality (HR=1.38; 95% CI: 1.07, 1.80), heart disease mortality (HR=1.58; 95% CI: 1.05, 2.51), and cancer mortality (HR=1.70; 95% CI: 1.08, 2.84). In single PFAS analysis, PFOS was found to be positively associated with all-cause mortality (third vs. first tertile HR=1.57; 95% CI: 1.22, 2.07), heart disease mortality (third vs. first tertile HR=1.65; 95% CI: 1.09, 2.57), and cancer mortality (third vs. first tertile HR=1.75; 95% CI: 1.10, 2.83), whereas PFOA exposure had no significant association with mortality. Assuming the observed association is causal, the number of deaths associated with PFOS exposure (≥17.1 vs. <7.9 ng/mL) was ∼382,000 (95% CI: 176,000, 588,000) annually between 1999 and 2015, and it decreased to 69,000 (95% CI: 28,000, 119,000) annually between 2015 and 2018. The association between PFOS and mortality was stronger among women and people without diabetes.

DISCUSSION

We observed a positive association between PFAS mixture exposure and mortality among U.S. adults. Limitations of this study include the potential for unmeasured confounding, selection bias, a relatively small number of deaths, and only measuring PFAS at one point in time. Further studies with serial measures of PFAS concentrations and longer follow-ups are necessary to elucidate the association between PFAS and mortality from specific causes. https://doi.org/10.1289/EHP10393.