Classification of patients with chronic disease by activation level using machine learning methods

Patient Activation Measure (PAM) measures the activation level of patients with chronic conditions and correlates well with patient adherence behavior, health outcomes, and healthcare costs. PAM is increasingly used in practice to identify patients needing more support from the care team. We define PAM levels 1 and 2 as low PAM and investigate the performance of eight machine learning methods (Logistic Regression, Lasso Regression, Ridge Regression, Random Forest, Gradient Boosted Trees, Support Vector Machines, Decision Trees, Neural Networks) to classify patients. Primary data collected from adult patients (n=431) with Diabetes Mellitus (DM) or Hypertension (HT) attending Family Health Centers in Istanbul, Turkey, is used to test the methods. [Formula: see text] of patients in the dataset have a low PAM level. Classification performance with several feature sets was analyzed to understand the relative importance of different types of information and provide insights. The most important features are found as whether the patient performs self-monitoring, smoking and exercise habits, education, and socio-economic status. The best performance was achieved with the Logistic Regression algorithm, with Area Under the Curve (AUC)=0.72 with the best performing feature set. Alternative feature sets with similar prediction performance are also presented. The prediction performance was inferior with an automated feature selection method, supporting the importance of using domain knowledge in machine learning.

Patient Activation for Self-Management in Patients with Idiopathic Pulmonary Fibrosis or Sarcoidosis

BACKGROUND

Self-management is considered important in the management of patients with idiopathic pulmonary fibrosis (IPF) or sarcoidosis. However, data about the degree of activation for self-management is lacking.

OBJECTIVES

The aim of the study was to determine the degree of activation for self-management in patients with IPF or sarcoidosis using the Patient Activation Measure (PAM) and to evaluate the association between PAM scores, clinical characteristics, and health-related outcomes.

STUDY DESIGN AND METHODS

This cross-sectional prospective study assessed besides the PAM also demographics, lung function, dyspnea (modified Medical Research Council [mMRC]), fatigue (Checklist Individual Strength-Fatigue [CIS-Fatigue]), anxiety/depression (Hospital Anxiety and Depression Scale [HADS-A/HADS-D]), and generic health status (EuroQol five-dimensional-five-level [EQ-5D-5L]).

RESULTS

Mean PAM was 55.0 (9.1) points in patients with IPF (n = 59) and low levels of patient activation for self-management (PAM ≤55.1 points) were present in 56% of the patients. PAM Scores correlated significantly (p < 0.05) with mMRC (ρ = -0.476), HADS-A (ρ = -0.326), HADS-D (ρ = -0.459), and EQ-5D-5L (ρ = 0.393). In patients with sarcoidosis (n = 59), the mean PAM score was 55.7 (11.0) points, and 46% of the patients reported low PAM levels. Significant correlations were found with mMRC (ρ = -0.356), HADS-A (ρ = -0.394), HADS-D (ρ = -0.478), and EQ-5D-5L (ρ = 0.313).

CONCLUSION

About half of the outpatients with IPF or sarcoidosis have a low degree of activation for self-management, and these patients generally report more dyspnea, anxiety, depression, and a lower health status. Whether patients with a low degree of activation can be successful in self-managing their disease remains to be determined.

Increasing patient activation scores of vulnerable youth by partnering medical residency programs with public high schools

OBJECTIVE

To assess whether participation in Stanford Youth Coaches Programs (SYCP) increases patient activation scores and patient activation levels for vulnerable youth from low income communities.

METHODS

From 2016 to 18, seven high schools and four residency programs in California, Alabama, Kansas and Missouri participated in SYCPs. Enrolled youth participants completed online pre and post-participation surveys including the Patient Activation Measure (PAM®10). We used paired T-tests, chi square tests, and linear multivariate models to compare pre-and post-scores and levels.

RESULTS

143 participants completed pre- and post-participation surveys. The PAM®10 mean pre-test score was 64.5 and post-test was 69.37, with mean difference 4.89 (p=.002). Participants showed significant improvement in patient activation levels after participation. 60 % participants in lowest activation Level 1; 63 % in Level 2; and 32 % in Level 3 moved to a higher level of activation after participation; 46 % who started in Level 4 moved down to Level 3 after participation.

CONCLUSION AND PRACTICE IMPLICATIONS

Participation in SYCPs has potential to significantly increase patient activation for vulnerable youth which could lead to lifelong improvements in health outcomes and decrease in healthcare costs.

Attendance, activation and health profiles of participants, a prospective study on a regional cardiometabolic disease self-management program in Laval, Canada

Background

Chronic diseases are responsible for over 70% of all deaths globally. While some self-management programs have been shown to be efficacious in preventing or altering trajectories for some chronic conditions, scaling-up and sustaining such programs beyond tightly-controlled study conditions remain a major challenge. CISSS-Laval partnered with the Cardiovascular Health Awareness Program team to co-develop Cible-santé/prévention and evaluate the first cohort of participants enrolled in the program, in order to better understand the program’s implementation and scope. The objective of the current study was to describe the profile of attendees and the level of engagement of participants in a new, region-wide cardiometabolic disease self-management program offered in Laval, Canada.

Methods

This was a prospective study with no comparison group. Potential participants were identified and referred to the program from April to December 2015 by their primary care health professional practicing in one of the city’s interdisciplinary primary care clinics. They had their blood pressure, waist circumference and body mass index measured by trained volunteers, and completed a questionnaire on health habits, level of activation and the risk of developing prediabetes and type 2 diabetes over the next 10 years.

Results

A descriptive analysis of the first cohort of 141 Cible-Santé/prévention participants showed very low attendance. Furthermore, only 1 in 10 of enrolled participants completed the full program. The program typically attracted adults with some risk factors associated with their conditions (high waist circumference, obesity), but with an already high level of knowledge, skills and confidence to participate in self-management activities.

Conclusion

This study provides a portrait of new participants to a self-management cardiometabolic disease program, which highlights the potential of supporting patients ready to make changes but also exposes the difficulty of attracting a larger number and diversity of participants and in encouraging completion of the program.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-021-10558-6.

Diabetes status modifies the efficacy of home-based kidney care for Zuni Indians in a randomized controlled trial

BACKGROUND

Home-Based Kidney Care (HBKC) is a pragmatic treatment approach that addresses patient preferences and cultural barriers to healthcare. We previously reported the results of a clinical trial of HBKC vs. usual care in a cohort of Zuni Indians in New Mexico. This study investigated the potential for differential efficacy of HBKC vs. usual care according to type 2 diabetes (T2DM) status.

METHODS

We analyzed the data from all individuals who participated in a randomized clinical trial that compared HBKC to usual care among patients with CKD, and assessed whether the effect of the HBKC intervention affected the subset of patients with T2DM differently than those individuals without T2DM. We used linear regression models to estimate the effect of HBKC on improvement in Patient Activation Measure (PAM) total scores within the groups of participants defined by T2DM status, and to compare the effects between these two groups. We used generalized estimating equations (GEE) to account for household clustering.

RESULTS

The original study enrolled 63 participants into the HBKC group, and 62 into the usual care. Ninety-eight of these individuals completed the 12-month intervention, 50 in the HBKC group and 48 in the usual care group. The present study compared the intervention effect in the 56 participants with T2DM (24 participants in the HBKC group and 32 in usual care) to the intervention effect in the 42 participants without T2DM (26 participants in the HBKC group and 16 in usual care). Those with T2DM who received the HBKC intervention experienced an average increase in PAM total scores of 16.0 points (95% Confidence Interval: 8.8-23.1) more than those with T2DM who were in the usual care group. For those without T2DM, the intervention had essentially no effect, with those who received the HBKC intervention having an average PAM total scores that was 1.4 points (95% C.I.: -12.4 to 9.6) lower than those who received usual care. There was a significantly different HBKC treatment effect by T2DM status (p = 0.02).

CONCLUSION

This secondary analysis suggests that the effectiveness of this HBKC intervention on increasing patient activation is most notable among those CKD patients who also have T2DM.

Validation of the patient activation measure in patients at discharge from hospitals and at distance from hospital care in Sweden

Background

The Patient Activation Measure (PAM) is a recognized measure of how active patients are in their care, and has been translated into several languages and cultural contexts. Patient activity, self-care, and health literacy have become increasingly important aspects of health care, and thus reliable measures of these are needed. However, a Swedish translation of PAM is currently lacking. The aim of the study was to translate and assess the validity and reliability of the Swedish PAM-13.

Methods

A self-report questionnaire was handed out to 521 patients at ten medical, geriatric, and surgical wards, and one Virtual Health Room. The Rasch model was employed, using the partial credit model, to assess the functioning of the PAM scale, item fit, targeting, unidimensionality, local independence, differential item functioning (DIF), and person-separation index. Evidence of substantive, content, structural, and external validity was examined.

Results

Of the 521 patients who were consecutively handed a questionnaire, 248 consented to participate, yielding a response rate of 47.6%. The average measure for each category advanced monotonically. The difficulty of the PAM items ranged from − 1.55 to 1.26. The infit and outfit values for the individual items were acceptable. Items 1, 2, and 4 showed disordered thresholds. The mean person location was 1.48 (SD = 1.66). The person-item map revealed that there were no item representations at the top of the scale. The evidence for unidimensionality was ambiguous and response dependency was seen in some items. DIF was found for age. The person separation index was 0.85.

Conclusion

The Swedish PAM-13 was reliable, but was not conclusively found to represent one underlying construct. It seems that the Swedish PAM-13 lacks strong evidence for substantive, content, and structural validity. Although valid and reliable measures of ability for activation in self-care among patients are highly warranted, we recommend further development of PAM-13 before application in everyday clinical care.

Low patient activation levels in frail older adults: a cross-sectional study

Background

Frail older adults are increasingly expected to self-manage their health and healthcare. We assessed the extent to which this group is able to take up this responsibility by measuring their level of activation as patients (i.e. their knowledge, skills and confidence to self-manage their health and healthcare). Further, we studied which characteristics of older adults were associated with patient activation.

Methods

In this cross-sectional study 200 frail, competent adults (median age 87 years) participated. Participants were community-dwelling adults who received home care and residents of care homes. Data were collected via personal interviews in participants’ homes. The main outcome measure was patient activation assessed by the short version of the Patient Activation Measure (PAM-13; range: 0–100). The PAM distinguishes four levels of increasing activation with level 1 indicating poor patient activation and level 4 adequate patient activation. Other studied variables were: multimorbidity, type of residency, frailty (Tilburg Frailty Index), mental competence (Mini Mental State Examination), health-related quality of life (SF-12), satisfaction with healthcare (subscale Patient Satisfaction Questionnaire) and personal characteristics (age, gender, marital status, educational level). Regression analyses were performed to investigate which variables were associated with patient activation.

Results

Participants had a median PAM score of 51. Thirty-nine percent had level 1 activation, 31% level 2, 26% level 3 and 5% level 4. Fifty-nine percent of community dwelling adults had level 1 or 2 activation versus 81% of care home residents (p = 0.007). Mental competence (Effect: 0.52, CI: 0.03–1.01, p = 0.04) and health-related quality of life (Effect: 0.15, CI: 0.01–0.30, p = 0.04 for physical health; Effect: 0.20, CI: 0.07–0.34, p = 0.003 for mental health) were positively associated with patient activation. Frailty (Effect: -1.06, CI: -1.75 – -0.36, p = 0.003) was negatively associated with patient activation.

Conclusions

The majority of this frail and very old study population, especially those with a lower health-related quality of life, may be unable to self-manage their health and healthcare to the level expected from them. The increasing population of frail older adults may need help in managing their health and healthcare.

Assessing Patient Activation among High-Need, High-Cost Patients in Urban Safety Net Care Settings

We sought to examine the literature using the Patient Activation Measure (PAM) or the Patient Enablement Instrument (PEI) with high-need, high-cost (HNHC) patients receiving care in urban safety net settings. Urban safety net care management programs serve low-income, racially/ethnically diverse patients living with multiple chronic conditions. Although many care management programs track patient progress with the PAM or the PEI, it is not clear whether the PAM or the PEI is an effective and appropriate tool for HNHC patients receiving care in urban safety net settings in the United States. We searched PubMed, EMBASE, Web of Science, and PsycINFO for articles published between 2004 and 2015 that used the PAM and between 1998 and 2015 that used the PEI. The search was limited to English-language articles conducted in the United States and published in peer-reviewed journals. To assess the utility of the PAM and the PEI in urban safety net care settings, we defined a HNHC patient sample as racially/ethnically diverse, low socioeconomic status (SES), and multimorbid. One hundred fourteen articles used the PAM. All articles using the PEI were conducted outside the U.S. and therefore were excluded. Nine PAM studies (8%) included participants similar to those receiving care in urban safety net settings, three of which were longitudinal. Two of the three longitudinal studies reported positive changes following interventions. Our results indicate that research on patient activation is not commonly conducted on racially and ethnically diverse, low SES, and multimorbid patients; therefore, there are few opportunities to assess the appropriateness of the PAM in such populations. Investigators expressed concerns with the potential unreliability and inappropriate nature of the PAM on multimorbid, older, and low-literacy patients. Thus, the PAM may not be able to accurately assess patient progress among HNHC patients receiving care in urban safety net settings. Assessing progress in the urban safety net care setting requires measures that account for the social and structural challenges and competing demands of HNHC patients.

Factors affecting patients' online health information-seeking behaviours: The role of the Patient Health Engagement (PHE) Model

OBJECTIVE

To identify the variables affecting patients' online health information-seeking behaviours by examining the relationships between patient participation in their healthcare and online health information-seeking behaviours.

METHODS

A cross-sectional survey of Italian chronic patients (N=352) was conducted on patient's online health information-seeking behaviours and patient participation-related variables. Structural equation modeling analysis was conducted to test the hypothesis.

RESULTS

This study showed how the healthcare professionals' ability to support chronic patients' autonomy affect patients' participation in their healthcare and patient's online health information-seeking behaviours. However, results do not confirm that the frequency of patients' online health-information seeking behavior has an impact on their adherence to medical prescriptions.

CONCLUSION

Assuming a psychosocial perspective, we have discussed how patients' engagement - conceived as the level of their emotional elaboration of the health condition - affects the patients' ability to search for and manage online health information.

PRACTICE IMPLICATION

To improve the effectiveness of patients' online health information-seeking behaviours and to enhance the effectiveness of technological interventions in this field, healthcare providers should target assessing and improving patient engagement and patient empowerment in their healthcare. It is important that health professionals acknowledge patients' online health information-seeking behaviours that they discuss the information offered by patients and guide them to reliable and accurate web sources.

Factors associated with patient activation in an older adult population with functional difficulties

OBJECTIVE

Patient activation, the patient's knowledge, skill, and confidence to manage his or her health, is an important indicator of future health and use of health care resources. Understanding factors associated with patient activation in an older population with functional difficulties may inform care in this population. This study aimed to determine whether patient activation is associated with depression, chronic conditions, family support, difficulties with activities of daily living (ADLs) and instrumental activities of daily living (IADLs), hospitalizations, education, and financial strain.

METHODS

(N=277), We administered surveys measuring patient activation, financial strain, depressive symptoms, family support, and chronic conditions to an older adult population. We tested association through multivariate linear regressions controlling for race, sex, and age.

RESULTS

Patient activation is significantly (p<0.05), positively associated with family support and self-rated overall health, and significantly (p<0.05), negatively associated with depressive symptoms and difficulties with ADLs and IADLs. We found no association between patient activation and financial stress, hospitalizations, and education.

CONCLUSIONS

Older age, depressive symptoms, and difficulties with ADLs and IADLs were associated with decreased patient activation.

PRACTICE IMPLICATIONS

Developing interventions tailored to older adults' level of patient activation has the potential to improve outcomes for this population.

The effect of a brief educational programme added to mental health treatment to improve patient activation: A randomized controlled trial in community mental health centres

OBJECTIVE

While there is growing interest in improving patient activation in general medical health services, there are too few randomized controlled trials in mental health settings which show how improvement can be achieved. Using the Patient Activation Measure-13 (PAM-13), we aimed to assess the effect of pre-treatment, peer co-led educational intervention on patient activation. Secondary outcomes included measures of patient satisfaction, well-being, mental health symptoms, motivation, and treatment participation.

METHODS

Patients from two community mental health centres were randomized to a control group (CG, n=26) receiving treatment as usual, or an intervention group (IG, n=26) consisting of a four-hour group educational seminar (aiming to encourage patients to adopt an active role in their treatment) followed by treatment as usual.

RESULTS

Only the IG improved on PAM-13, at one- and four-month follow-ups. The intervention had significant effects on patient satisfaction and treatment participation, compared to CG.

CONCLUSION

Providing pre-treatment, peer co-led education improves patient activation in community mental health care settings.

PRACTICE IMPLICATIONS

The use of peers as co-educators may contribute to a different mental health care delivery, ensuring patient activation and participation in treatment. Further studies should examine peers' needs for supervision, challenges for the services, long-term and cost-benefit effects. CLINICALTRIALS.

GOV IDENTIFIER

NCT01601587.