Using existing questionnaires in latent class analysis: should we use summary scores or single items as input? A methodological study using a cohort of patients with low back pain

Associations Among Lifetime Discrimination Typologies and Psychological Health in Black and Hispanic Women After Birth

OBJECTIVE

To identify lifetime discrimination typologies and examine their associations with psychological health outcomes among Black and Hispanic women after birth.

DESIGN

Secondary analysis of the Community and Child Health Network study data.

PARTICIPANTS

A total of 1,350 Black and 607 Hispanic women.

METHODS

We built two latent class models for Black and Hispanic women using eight indicator variables from different life domains of discrimination (childhood, family, work, police, education, housing, health care, and loans). We used bivariate and multiple regression analyses to examine the association among the identified typologies and postpartum depression and perceived stress at 6 months postpartum.

RESULTS

We selected the three-class model with best fit indices and interpretability: no lifetime discrimination (n = 1,029; 76.22%), high childhood-family racial discrimination (n = 224, 16.59%), and moderate lifetime discrimination (n = 97, 7.19%) among Black women and no lifetime discrimination (n = 493, 81.22%), high childhood-family racial discrimination (n = 93, 15.32%), and high education discrimination (n = 21, 3.46%) among Hispanic women. The adjusted postpartum depression and perceived stress scores were significantly greater in Typologies 2 and 3 than Typology 1 in Black women. The adjusted perceived stress scores were significantly greater in Typologies 2 and 3 than Typology 1 in Hispanic women.

CONCLUSION

Lifetime discrimination experiences manifested in complex patterns. Women who experienced moderate to high discrimination across all or specific life domains had worse postpartum depression and perceived stress at 6 months after birth. It is crucial to address lifetime discrimination to improve maternal mental health.

Mental health outcomes of physical, sexual, and psychological intimate partner violence among women in Turkey: A latent class study

Intimate partner violence (IPV) is a severe human right violation and a significant public health problem in Turkey. However, its complexity and stratified relationship with mental health problems are still uncovered in this under researched community, as violence is often approached as a uniform health and social problem. In this study, we collected data from 539 female adults in heterosexual relationships in Turkey using convenience and snowball sampling. The aim was to examine violence-related clusters and their association of these clusters with mental health outcomes, including depression, anxiety, and depression). We assessed whether participants experienced physical, sexual, and psychological violence in their intimate relationships and conducted a latent class analysis. We obtained five violence clusters: (1) no or low violence, (2) high psychological violence, (3) jealous control with physical violence, (4) moderate sexual and psychological violence, and (5) severe physical, sexual, and psychological violence. Analyses of variance showed that women who experienced multiple forms of IPV reported stronger depression, anxiety, and somatization compared with women in the no or low violence class. Additionally, they reported greater depression and anxiety compared with women who experienced only high psychological IPV. However, psychological violence amplified somatization as much as physical and sexual violence.

Fathering styles in a traditional culture and its association with marital relationship: A latent profile analysis with a nationally representative sample

This study had two main objectives. The first goal was to examine fathers' parenting styles in a society with traditional patriarchal values. The second goal was to understand the extent to which marital relationship variables (i.e., marital satisfaction and spousal support) are related with the identified fathering profiles. The theoretical and empirical underpinnings of the "parenting styles" literature that form the foundation of the research on parent-child dynamics are still mainly based on data collected only from mothers. This weakness of the literature on fathering typologies is further accompanied by the scarcity of data on fathers' parenting practices from non-Western contexts. To achieve the study aims, a latent profile analysis based on warmth and control dimensions of parenting was performed using a nationally representative sample of 1070 urban fathers in Turkey. Using scores on six parenting behavior variables (i.e., warmth, punishment, inductive reasoning, positive parenting, discipline and teaching responsibilities, time and talking together), results revealed four fathering profiles: "authoritative" (30%), "average" (45%), "uninvolved" (16%), and "disciplining-distant" (8%). Furthermore, the results showed that the fathering profiles significantly differed from each other on the marital satisfaction and spousal support scores. Results highlight the importance of investigating fathering behaviors and their predictors in different cultures.

Identifying chronic low back pain phenotypic domains and characteristics accounting for individual variation: a systematic review

ABSTRACT

Interpatient variability is frequently observed among individuals with chronic low back pain (cLBP). This review aimed at identifying phenotypic domains and characteristics that account for interpatient variability in cLBP. We searched MEDLINE ALL (through Ovid), Embase Classic and EMBASE (through Ovid), Scopus, and CINAHL Complete (through EBSCOhost) databases. Studies that aimed to identify or predict cLBP different phenotypes were included. We excluded studies that focused on specific treatments. The methodological quality was assessed using an adaptation of the Downs and Black tool. Forty-three studies were included. Although the patient and pain-related characteristics used to identify phenotypes varied considerably across studies, the following were among the most identified phenotypic domains and characteristics that account for interpatient variability in cLBP: pain-related characteristics (including location, severity, qualities, and duration) and pain impact (including disability, sleep, and fatigue), psychological domains (including anxiety and depression), behavioral domains (including coping, somatization, fear avoidance, and catastrophizing), social domains (including employment and social support), and sensory profiling (including pain sensitivity and sensitization). Despite these findings, our review showed that the evidence on pain phenotyping still requires further investigation. The assessment of the methodological quality revealed several limitations. We recommend adopting a standard methodology to enhance the generalizability of the results and the implementation of a comprehensive and feasible assessment framework to facilitate personalized treatments in clinical settings.

Systems Medicine Approach for Tinnitus with Comorbid Disorders

Despite the fact that chronic diseases usually occur together with a spectrum of possible comorbidities that may differ strongly between patients, they are classically still viewed as distinct disease entities and, consequently, are often treated with uniform therapies. Unfortunately, such an approach does not take into account that different combinations of symptoms and comorbidities may result from different pathological (e.g., environmental, genetic, dietary, etc.) factors, which require specific and individualised therapeutic strategies. In this opinion paper, we aim to put forward a more differentiated, systems medicine approach to disease and patient treatment. To elaborate on this concept, we focus on the interplay of tinnitus, depression, and chronic pain. In our view, these conditions can be characterised by a variety of phenotypes composed of variable sets of symptoms and biomarkers, rather than distinct disease entities. The knowledge of the interplay of such symptoms and biomarkers will provide the key to a deeper, mechanistic understanding of disease pathologies. This paves the way for prediction and prevention of disease pathways, including more personalised and effective treatment strategies.

Gender differences in the experience of interstitial cystitis/bladder pain syndrome

Aims

This study assessed gender differences in a debilitating urologic pain condition, interstitial cystitis/bladder pain syndrome (IC/BPS). We aimed to (1) evaluate how pain, symptom, and distress profiles of IC/BPS may differ between genders and (2) obtain in-depth firsthand accounts from patients to provide additional insight into their experiences that may explain potential gender differences.

Methods

A mixed methods approach combined validated patient-reported outcome measures with a single timepoint 90-min focus group. Tests of summary score group differences between men and women were assessed across questionnaires measuring urologic symptoms, pain, emotional functioning, and diagnostic timeline. Qualitative analysis applied an inductive-deductive approach to evaluate and compare experiences of living with IC/BPS Group narratives were coded and evaluated thematically by gender using the biopsychosocial model, providing insight into the different context of biopsychosocial domains characterizing the male and female experience of IC/BPS.

Results

Thirty-seven participants [women (n = 27) and men (n = 10)] completed measures and structured focus group interviews across eight group cohorts conducted from 8/2017 to 3/2019. Women reported greater pain intensity (p = 0.043) and extent (p = 0.018), but not significantly greater impairment from pain (p = 0.160). Levels of psychological distress were significantly elevated across both genders. Further, the duration between time of pain symptom onset and time to diagnosis was significantly greater for women than men (p = 0.012). Qualitative findings demonstrated key distinctions in experiences between genders. Men appeared not to recognize or to deter emotional distress while women felt overwhelmed by it. Men emphasized needing more physiological treatment options whilst women emphasized needing more social and emotional support. Interactions with medical providers and the healthcare system differed substantially between genders. While men reported feeling supported and involved in treatment decisions, women reported feeling dismissed and disbelieved.

Conclusion

The findings indicate different pain experiences and treatment needs between genders in persons experiencing urologic pain and urinary symptoms, with potential intervention implications. Results suggest gender health inequality in medical interactions in this urologic population needing further investigation.

Differential Adaptation to Adversity: A Latent Profile Analysis of Youth Engagement With Resilience-Enabling Cultural Resources and Mental Health Outcomes in a Stressed Canadian and South African Community

Using person-centered latent profile analyses, this article reports two distinct sub-groups—nominal versus robust cultural allegiance—that characterize how a sample of 14- to 24-year-olds from stressed environments in South Africa ( n = 576, nfemales = 314, nmales = 257) and Canada ( n =V481; nfemales = 270, nmales = 211) engage with four cultural resources (spirituality, religiosity, family tradition, and community tradition). It considers how nominal versus robust cultural allegiance is associated with youths’ self-reported symptoms of depression and conduct disorder, age-group, and gender. In doing so, the article addresses pre-existing resilience studies’ general inattention to patterns of differential adaptation in how specific groups of youth adjust to adversity, and the role of cultural resources in youth mental health. The results draw attention to the importance of understanding resilience in sociocultural context and urge mental health practitioners and other resilience champions to be circumspect in their work with at-risk youth about which cultural resources they leverage for which mental health outcomes.

How does individualised physiotherapy work for people with low back pain? A Bayesian Network analysis using randomised controlled trial data

PURPOSE

Individualised physiotherapy is an effective treatment for low back pain. We sought to determine how this treatment works by using randomised controlled trial data to develop a Bayesian Network model.

METHODS

300 randomised controlled trial participants (153 male, 147 female, mean age 44.1) with low back pain (of duration 6-26 weeks) received either individualised physiotherapy or advice. Variables with potential to explain how individualised physiotherapy works were included in a multivariate Bayesian Network model. Modelling incorporated the intervention period (0-10 weeks after study commencement-"early" changes) and the follow-up period (10-52 weeks after study commencement-"late" changes). Sequences of variables in the Bayesian Network showed the most common direct and indirect recovery pathways followed by participants with low back pain receiving individualised physiotherapy versus advice.

RESULTS

Individualised physiotherapy directly reduced early disability in people with low back pain. Individualised physiotherapy exerted indirect effects on pain intensity, recovery expectations, sleep, fear, anxiety, and depression via its ability to facilitate early improvement in disability. Early improvement in disability, led to an early reduction in depression both directly and via more complex pathways involving fear, recovery expectations, anxiety, and pain intensity. Individualised physiotherapy had its greatest influence on early change variables (during the intervention period).

CONCLUSION

Individualised physiotherapy for low back pain appears to work predominately by facilitating an early reduction in disability, which in turn leads to improvements in other biopsychosocial outcomes. The current study cannot rule out that unmeasured mechanisms (such as tissue healing or reduced inflammation) may mediate the relationship between individualised physiotherapy treatment and improvement in disability. Further data-driven analyses involving a broad range of plausible biopsychosocial variables are recommended to fully understand how treatments work for people with low back pain.

TRIALS REGISTRATION

ACTRN12609000834257.

Machine learning approaches applied in spinal pain research

The purpose of this narrative review is to provide a critical reflection of how analytical machine learning approaches could provide the platform to harness variability of patient presentation to enhance clinical prediction. The review includes a summary of current knowledge on the physiological adaptations present in people with spinal pain. We discuss how contemporary evidence highlights the importance of not relying on single features when characterizing patients given the variability of physiological adaptations present in people with spinal pain. The advantages and disadvantages of current analytical strategies in contemporary basic science and epidemiological research are reviewed and we consider how analytical machine learning approaches could provide the platform to harness the variability of patient presentations to enhance clinical prediction of pain persistence or recurrence. We propose that machine learning techniques can be leveraged to translate a potentially heterogeneous set of variables into clinically useful information with the potential to enhance patient management.

Montreal Cognitive Assessment: Seeking a Single Cutoff Score May Not Be Optimal

Background

Cutoff scores of the Montreal cognitive assessment (MoCA) for screening mild cognitive impairment in older adults differ across the world and within the Chinese culture. It is argued that to seek a cutoff score is essential to classify test participants. It was unknown how taking a classifying approach might reveal the cutoff score for identifying mildly cognitively impaired older adults.

Methods

Participants, selected from 13 communities in Wuhan, China, were tested with the Chinese version of MoCA and rated with the Activities of Daily Living and the Clinical Dementia Rating scales. Mixture modeling was applied to the data with certain covariates and MoCA sum scores as the outcome of the latent class. Models with different numbers of classes were compared in terms of information criteria, likelihood ratio test, entropy, and interpretability.

Results

A 3-class model (normal, mildly impaired, and severely impaired) was found to fit the data best. The normal class averaged a MoCA score of 24, while the severely impaired class averaged a score below 18. For those cases with MoCA scores above 18 and below 24, it is not certain if they are in the normal or the severely impaired classes.

Conclusion

Latent variable classification modeling provides another option to identify MCI in older adults. Some categorically different cases of MCI cannot be captured with any single MoCA sum score. A range of 18–24 MoCA scores might serve as a better screening criterion of MCI. Older adults who scored within this gray zone should be monitored for potential interventions.

Assessment of Common Comorbidity Phenotypes Among Older Adults With Knee Osteoarthritis to Inform Integrated Care Models

Objective

To establish the frequency of concordant, discordant, and clinically dominant comorbidities among Medicare beneficiaries with knee osteoarthritis (KOA) and to identify common concordant condition subgroups.

Participants and Methods

We used a 5% representative sample of Medicare claims data to identify beneficiaries who received a diagnosis of KOA between January 1, 2012, and September 30, 2015, and matched control group without an osteoarthritis (OA) diagnosis. Frequency of 34 comorbid conditions was categorized as concordant, discordant, or clinically dominant among those with KOA and a matched sample without OA. Comorbid condition phenotypes were characterized by concordant conditions and derived using latent class analysis among those with KOA.

Results

The study sample included 203,361 beneficiaries with KOA and 203,361 non-OA controls. The largest difference in frequency between the two cohorts was for co-occurring musculoskeletal conditions (23.7% absolute difference), chronic pain syndromes (6.5%), and rheumatic diseases (4.5%), all with a higher frequency among those with knee OA. Phenotypes were identified as low comorbidity (53% of cohort with classification), hypothyroid/osteoporosis (27%), vascular disease (10%), and high medical and psychological comorbidity (10%).

Conclusions

Approximately 47% of Medicare beneficiaries with KOA in this sample had a phenotype characterized by one or more concordant conditions, suggesting that existing clinical pathways that rely on single or dominant providers might be insufficient for a large proportion of older adults with KOA. These findings could guide development of integrated KOA-comorbidity care pathways that are responsive to emerging priorities for personalized, value-based health care.

Alcohol-Related Risky Behavior Patterns and Their Association With Alcohol Use and Perceived Alcohol Stigma in Moshi, Tanzania

OBJECTIVE

The Kilimanjaro region has one of the highest rates of reported alcohol use per capita in Tanzania. Alcohol-related risky behaviors pose substantial threats to the health and well-being of alcohol users and the people around them. This study seeks to understand how alcohol-related risky behaviors co-occur with other risky behaviors.

METHOD

Latent class analysis (LCA) was applied to examine alcohol-related risky behaviors. The optimal number of latent classes was confirmed by using model fit indices. Negative binomial models were used to test latent classes and their association with harmful and hazardous drinking and perceived alcohol stigma. With the model defined, we explored each class's drinking patterns and risky behavior patterns.

RESULTS

A total of 622 (60% male) of 841 participants were included in these analyses because they drank alcohol at least once in their lifetime. Three classes of risky behavior patterns were identified: Class 1, "Limited risk behaviors" (59.7%); Class 2, "Primarily foolish behaviors" (25.6%); and Class 3, "Pervasive risk behaviors" (13.1%). Class 3 had the most alcohol use quantity and frequency. No association between classes and alcohol stigma was found. Compared with males, females are less likely to be classified in Class 2 and 3.

CONCLUSIONS

Three different classes of risky behaviors became apparent and were distinguished by gender, age, and personal alcohol use. Our findings suggest a potential role for personalized interventions based on latent classes specifically to reduce risk behaviors.

Alcohol-Related Risky Behavior Patterns and Their Association With Alcohol Use and Perceived Alcohol Stigma in Moshi, Tanzania

OBJECTIVE

The Kilimanjaro region has one of the highest rates of reported alcohol use per capita in Tanzania. Alcohol-related risky behaviors pose substantial threats to the health and well-being of alcohol users and the people around them. This study seeks to understand how alcohol-related risky behaviors co-occur with other risky behaviors.

METHOD

Latent class analysis (LCA) was applied to examine alcohol-related risky behaviors. The optimal number of latent classes was confirmed by using model fit indices. Negative binomial models were used to test latent classes and their association with harmful and hazardous drinking and perceived alcohol stigma. With the model defined, we explored each class's drinking patterns and risky behavior patterns.

RESULTS

A total of 622 (60% male) of 841 participants were included in these analyses because they drank alcohol at least once in their lifetime. Three classes of risky behavior patterns were identified: Class 1, "Limited risk behaviors" (59.7%); Class 2, "Primarily foolish behaviors" (25.6%); and Class 3, "Pervasive risk behaviors" (13.1%). Class 3 had the most alcohol use quantity and frequency. No association between classes and alcohol stigma was found. Compared with males, females are less likely to be classified in Class 2 and 3.

CONCLUSIONS

Three different classes of risky behaviors became apparent and were distinguished by gender, age, and personal alcohol use. Our findings suggest a potential role for personalized interventions based on latent classes specifically to reduce risk behaviors.

Comorbidity Subgroups Among Medicare Beneficiaries Seeking Health Care for Musculoskeletal Pain

BACKGROUND

Treatment of musculoskeletal pain in older adults may be more effective if it incorporates integrated management of comorbid health conditions. The purpose of this study was to define empirically derived comorbidity subgroups among Medicare beneficiaries with an index condition of osteoarthritis (OA) or low back pain (LBP) as a precursor to the development of comorbidity-specific pain treatment pathways.

METHODS

This study included Medicare beneficiaries participating in the Medicare Current Beneficiary Survey (MCBS) and seeking care for OA (n = 723) or LBP (n = 617) with data available for 3 years after entry into the survey. We identified 30 comorbidity diagnoses using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnostic codes in claims data during beneficiaries' first year in the survey. Latent class analysis defined comorbidity subgroups, and posterior probabilities were used to assign subgroup classification. Self-reported disability was compared over three consecutive years for each subgroup.

RESULTS

We identified similar comorbidity subgroups for OA and LBP. The subgroups included (range of percent prevalence) low comorbidity (47.6%-54.4%), nonvascular (21.8%-28.6%), diabetes (12.2%-15.0%), renal disease with complicated hypertension (5.5%-5.8%), and complex cardiac disease/high comorbidity (3.3%-5.8%). OA and LBP subgroups with more complex comorbidity burden generally demonstrated higher disability over 3 years.

CONCLUSIONS

Five comorbidity subgroups were identified, with a large proportion of older adults classified into the subgroup defined by a low probability of most comorbidities. These findings provide direction for the development of pain treatment pathways that are tailored to address common comorbidity profiles among older adults.

Exploring conceptual preprocessing for developing prognostic models: a case study in low back pain patients

OBJECTIVES

A conceptually oriented preprocessing of a large number of potential prognostic factors may improve the development of a prognostic model. This study investigated whether various forms of conceptually oriented preprocessing or the preselection of established factors was superior to using all factors as input.

STUDY DESIGN AND SETTING

We made use of an existing project that developed two conceptually oriented subgroupings of low back pain patients. Based on the prediction of six outcome variables by seven statistical methods, this type of preprocessing was compared with medical experts' preselection of established factors, as well as using all 112 available baseline factors.

RESULTS

Subgrouping of patients was associated with low prognostic capacity. Applying a Lasso-based variable selection to all factors or to domain-specific principal component scores performed best. The preselection of established factors showed a good compromise between model complexity and prognostic capacity.

CONCLUSION

The prognostic capacity is hard to improve by means of a conceptually oriented preprocessing when compared to purely statistical approaches. However, a careful selection of already established factors combined in a simple linear model should be considered as an option when constructing a new prognostic rule based on a large number of potential prognostic factors.

Chronic low back pain is highly individualised: patterns of classification across three unidimensional subgrouping analyses

BACKGROUND AND AIMS

Chronic low back pain (CLBP) is a complex disorder where central and peripheral nociceptive processes are influenced by factors from multiple dimensions associated with CLBP (e.g. movement, pain sensitivity, psychological). To date, outcomes for treatments matched to unidimensional subgroups (e.g. psychologically-based) have been poor. Therefore, unidimensional subgrouping may not reflect the complexity of CLBP presentations at an individual level. The aim of this study was therefore to explore patterns of classification at an individual level across the three previously-published, data-driven, within-dimension subgrouping studies.

METHODS

Cross-sectional, multidimensional data was collected in 294 people with CLBP. Statistical derivation of subgroups within each of three clinically-important dimensions (pain sensitivity, psychological profile, pain responses following repeated spinal bending) was briefly reviewed. Patterns of classification membership were subsequently tabulated across the three dimensions.

RESULTS

Of 27 possible patterns across these dimensions, 26 were represented across the cohort.

CONCLUSIONS

This result highlights that while unidimensional subgrouping has been thought useful to guide treatment, it is unlikely to capture the full complexity of CLBP. The amount of complexity important for best patient outcomes is currently untested.

IMPLICATIONS

For clinicians this study highlights the high variability of presentations of people with CLBP at the level of the individual. For example, clinician's should not assume that those with high levels of pain sensitivity will also have high psychological distress and have pain summation following repeated spinal bending. A more flexible, multidimensional, clinically-reasoned approach to profile patient complexity may be required to inform individualised, patient-centred care. Such individualised care might improve treatment efficacy. This study also has implications for researchers; highlighting the inadequacy of unidimensional subgrouping processes and methodological difficulties in deriving subgroups across multidimensional data.

Typologies of PTSD clusters and reckless/self-destructive behaviors: A latent profile analysis

Posttraumatic stress disorder (PTSD) is comorbid with diverse reckless and self-destructive behaviors (RSDBs). We examined the nature and construct validity (covariates of age, gender, depression severity, number of trauma types, functional impairment) of the optimal class solution categorizing participants based on PTSD symptom and RSDB endorsement. The sample included 417 trauma-exposed individuals recruited through Amazon's MTurk platform who completed the Life Events Checklist for DSM-5, PTSD Checklist for DSM-5, the Posttrauma Risky Behaviors Questionnaire, and Patient Health Questionnaire-9. Latent profile analyses indicated an optimal three-class solution: the Low PTSD-RSDBs, High PTSD-Low RSDBs, and High PTSD-RSDBs classes. Multinomial logistic regression indicated that impairment and depression predicted the High PTSD-Low RSDBs vs. the Low PTSD-RSDBs classes. Impairment, age, being female, and depression predicted the High vs. Low PTSD-RSDBs classes. Number of trauma types, age, being female, and depression predicted the High PTSD-RSDBs vs. High PTSD-Low RSDBs classes. Results support the presence of a reckless behaviors subtype of PTSD (characterized by greater depression, greater impariment, greater number of trauma types, being male, and being younger), conducting comprehensive assessments of RSDBs for individuals reporting PTSD symptoms and of PTSD symptoms for individuals reporting RSDBs, and the need to tailor interventions to treat PTSD and RSDBs concurrently.

Using item response theory with health system data to identify latent groups of patients with multiple health conditions

A critical step toward tailoring effective interventions for heterogeneous and medically complex patients is to identify clinically meaningful subgroups on the basis of their comorbid conditions. We applied Item Response Theory (IRT), a potentially useful tool to identify clinically meaningful subgroups, to characterize phenotypes within a cohort of high-risk patients. This was a retrospective cohort study using 68,400 high-risk Veteran’s Health Administration (VHA) patients. Thirty-one physical and mental health diagnosis indicators based on ICD-9 codes from patients’ inpatient, outpatient VHA and VA-paid community care claims. Results revealed 6 distinct subgroups of high-risk patients were identified: substance use, complex mental health, complex diabetes, liver disease, cancer with cardiovascular disease, and cancer with mental health. Multinomial analyses showed that subgroups significantly differed on demographic and utilization variables which underscored the uniqueness of the groups. Using IRT models with clinical diagnoses from electronic health records permitted identification of diagnostic constellations among otherwise undifferentiated high-risk patients. Recognizing distinct patient profiles provides a framework from which insights into medical complexity of high-risk patients can be explored and effective interventions can be tailored.

Latent class analysis derived subgroups of low back pain patients - do they have prognostic capacity?

BACKGROUND

Heterogeneity in patients with low back pain is well recognised and different approaches to subgrouping have been proposed. One statistical technique that is increasingly being used is Latent Class Analysis as it performs subgrouping based on pattern recognition with high accuracy. Previously, we developed two novel suggestions for subgrouping patients with low back pain based on Latent Class Analysis of patient baseline characteristics (patient history and physical examination), which resulted in 7 subgroups when using a single-stage analysis, and 9 subgroups when using a two-stage approach. However, their prognostic capacity was unexplored. This study (i) determined whether the subgrouping approaches were associated with the future outcomes of pain intensity, pain frequency and disability, (ii) assessed whether one of these two approaches was more strongly or more consistently associated with these outcomes, and (iii) assessed the performance of the novel subgroupings as compared to the following variables: two existing subgrouping tools (STarT Back Tool and Quebec Task Force classification), four baseline characteristics and a group of previously identified domain-specific patient categorisations (collectively, the 'comparator variables').

METHODS

This was a longitudinal cohort study of 928 patients consulting for low back pain in primary care. The associations between each subgroup approach and outcomes at 2 weeks, 3 and 12 months, and with weekly SMS responses were tested in linear regression models, and their prognostic capacity (variance explained) was compared to that of the comparator variables listed above.

RESULTS

The two previously identified subgroupings were similarly associated with all outcomes. The prognostic capacity of both subgroupings was better than that of the comparator variables, except for participants' recovery beliefs and the domain-specific categorisations, but was still limited. The explained variance ranged from 4.3%-6.9% for pain intensity and from 6.8%-20.3% for disability, and highest at the 2 weeks follow-up.

CONCLUSIONS

Latent Class-derived subgroups provided additional prognostic information when compared to a range of variables, but the improvements were not substantial enough to warrant further development into a new prognostic tool. Further research could investigate if these novel subgrouping approaches may help to improve existing tools that subgroup low back pain patients.

Identifying subgroups of patients using latent class analysis: should we use a single-stage or a two-stage approach? A methodological study using a cohort of patients with low back pain

Background

Heterogeneity in patients with low back pain (LBP) is well recognised and different approaches to subgrouping have been proposed. Latent Class Analysis (LCA) is a statistical technique that is increasingly being used to identify subgroups based on patient characteristics. However, as LBP is a complex multi-domain condition, the optimal approach when using LCA is unknown. Therefore, this paper describes the exploration of two approaches to LCA that may help improve the identification of clinically relevant and interpretable LBP subgroups.

Methods

From 928 LBP patients consulting a chiropractor, baseline data were used as input to the statistical subgrouping. In a single-stage LCA, all variables were modelled simultaneously to identify patient subgroups. In a two-stage LCA, we used the latent class membership from our previously published LCA within each of six domains of health (activity, contextual factors, pain, participation, physical impairment and psychology) (first stage) as the variables entered into the second stage of the two-stage LCA to identify patient subgroups. The description of the results of the single-stage and two-stage LCA was based on a combination of statistical performance measures, qualitative evaluation of clinical interpretability (face validity) and a subgroup membership comparison.

Results

For the single-stage LCA, a model solution with seven patient subgroups was preferred, and for the two-stage LCA, a nine patient subgroup model. Both approaches identified similar, but not identical, patient subgroups characterised by (i) mild intermittent LBP, (ii) recent severe LBP and activity limitations, (iii) very recent severe LBP with both activity and participation limitations, (iv) work-related LBP, (v) LBP and several negative consequences and (vi) LBP with nerve root involvement.

Conclusions

Both approaches identified clinically interpretable patient subgroups. The potential importance of these subgroups needs to be investigated by exploring whether they can be identified in other cohorts and by examining their possible association with patient outcomes. This may inform the selection of a preferred LCA approach.

Electronic supplementary material

The online version of this article (doi:10.1186/s12891-017-1411-x) contains supplementary material, which is available to authorized users.