Reliability and Validity of the Patient Activation Measure in Kidney Disease: Results of Rasch Analysis

Abstract

Visual Abstract

Background and objectives 

Despite the increasing prioritization of the promotion of patient activation in nephrology, its applicability to people with CKD is not well established. Before the Patient Activation Measure is universally adopted for use in CKD, it is important to critically evaluate this measure. The aim of this study was to describe the psychometric properties of the Patient Activation Measure in CKD.

Design, setting, participants, & measurements 

A survey containing the 13-item Patient Activation Measure was completed by 942 patients with CKD, not treated with dialysis. Data quality was assessed by mean, item response, missing values, floor and ceiling effects, internal consistency (Cronbach’s alpha and average interitem correlation), and item-rest correlations. Rasch modeling was used to assess item performance and scaling (item statistics, person and item reliability, rating scale diagnostics, factorial test of residuals, and differential item functioning).

Results 

The item response was high, with a small number of missing values (<1%). Floor effect was small (range 1%–5%), but the ceiling effect was above 15% for nine items (range 15%–38%). The Patient Activation Measure demonstrated good internal consistency overall (Cronbach α=0.925, and average interitem correlation 0.502). The difficulty of the Patient Activation Measure items ranged from −0.90 to 0.86. Differential item functioning was found for disease type (item 3) and age (item 12). The person separation index was 9.48 and item separation index was 3.21.

Conclusions 

The 13-item Patient Activation Measure appears to be a suitably reliable and valid instrument for assessing patient activation in CKD. In the absence of a kidney-specific instrument, our results support the 13-item Patient Activation Measure as a promising measure to assess activation in those with CKD, although consideration for several items is warranted. The high ceiling effect may be a problem when using the 13-item Patient Activation Measure to measure changes over time.

Introduction

Patient activation describes the knowledge, skills, and confidence in managing one’s own health (¹). Patient activation is linked to improved health outcomes in chronic conditions, such as hypertension, diabetes, and heart disease (²,³); in patients with CKD, lower activation levels are associated with older age, receiving in-center hemodialysis, poorer perceptions of health-related quality of life, higher decisional conflict regarding treatment/modality choice, and lower medication adherence (⁴). Engaging people in their own health care is a growing focus of many health care models, particularly in long-term conditions (⁵).

The most frequently used measure of patient activation is the 13-item Patient Activation Measure (PAM-13), a self-report tool designed to assess one’s activation in managing one's health (⁶). The PAM scores can be used to define respondents into one of four activation levels, from passive and lacking knowledge and skills (Level 1) to active, well informed, and competent (Level 4) (⁷,⁸). Interest in the PAM-13 as a tool to facilitate the delivery of person-centered care across the United Kingdom (UK) is growing (⁹). The concept of patient activation, and its measurement as an indicator of quality and effectiveness, is receiving increasing attention from the National Health Service (NHS) in its provision of person-centered, long-term condition management (⁵). The PAM-13 was piloted in the NHS, through the UK Renal Registry, as an outcome measure in the Valuing Individuals: Transforming Participation in CKD program (¹⁰). It is also endorsed by the National Quality Forum Quality Positioning System and the Kidney Care Quality Alliance. In the United States, patient activation was recently included in two options of the Kidney Care Choices payment model of the Centers for Medicare and Medicaid Services: the Kidney Care First and the Comprehensive Kidney Care Contracting models (⁴).

Translated into multiple languages, the PAM-13 has been validated and utilized in many regions, including the United States (¹), Europe (^{1112131415–16}), and Asia (¹⁷,¹⁸), and across different clinical settings and multiple patient populations (^{1920212223–24}). Despite the increasing prioritization of patient activation in nephrology (⁴), its applicability to people with CKD is not well established, and no research has determined the psychometric properties of the PAM-13 in a CKD population. It remains to be determined whether the PAM-13 accurately captures the construct of patient activation (⁴), and before the PAM-13 is universally recognized for use in CKD, it is important to critically evaluate this measure as it applies to CKD. The purpose of this paper is to describe the psychometric properties of the PAM-13 in CKD.

Materials and Methods

Study Design

This analysis consists of data taken from a multicenter observational study (DIMENSION-KD, ISRCTN84422148) where participants completed a self-administered survey pack comprising questionnaires designed to assess lifestyle factors (e.g., physical function and activity). Completed at home, the survey pack was returned to the research team via the post. Participants were recruited from nephrology outpatient clinics and from primary care practices between July 2018 and February 2020, across 14 sites in England.

Patient Population

Participants were included if they were: (1) diagnosed with a kidney condition (CKD <60 ml/min per 1.73 m² not requiring dialysis, and kidney transplant recipients), (2) aged ≥18 years, and (3) able to provide informed consent. Participants undergoing dialysis treatment were excluded. The study was granted national research ethical approval by the Leicester Research Ethics Committee (18/EM/0117). All patients provided written informed consent, and the study was conducted in accordance with the Declaration of Helsinki. Demographic and clinical data were collected from a composite of self-reported information and electronic medical records.

Patient Activation Measure

The PAM-13 (¹) is the short form of the 22-item PAM (⁶), measuring the knowledge, skills, and confidence for self-management. Individuals respond to items (e.g., “I know how to prevent further problems with my health condition”) using a four-point Likert scale ranging from “strongly disagree” (¹) to “strongly agree” (⁴). A “not applicable” (N/A) response is also available. Responses of N/A are scored as 0 and are reported to distinguish from those left blank (²⁵). A continuous activation score is computed from the raw score using an empirically derived calibration table by Insignia Health. The PAM-13 is scored along a Guttman scale (0–100), with higher scores along a unidimensional continuum signifying greater activation. Level 1 (PAM-13 score ≤47; disengagement and disbelief about one’s own role in self-management) encompasses items 1 and 2; Level 2 (47.1–55.1; increasing awareness, confidence, and knowledge), 3–8; Level 3 (55.2–67; readiness and taking action), 9–11; and Level 4 (≥67.1; sustainment), 12 and 13 (Figure 1). The PAM-13 has information-weighted fit/outlier-sensitive fit (INFIT/OUTFIT) statistics ranging from 0.50 to 1.50 and a Cronbach’s alpha of 0.88.

Figure 1.: Description of activation level of the Patient Activation Measure.

Psychometric Analysis

Data Quality and Item Analysis.

Basic item descriptives and data quality were assessed in terms of item mean with SD, missing data, number of N/A answers, skewness with SEM, and extent of ceiling and floor effects (defined as >15% [²⁶]). Invalid questionnaires with ≥3 missing items were excluded.

Internal Consistency (Reliability).

Internal consistency was assessed using Cronbach’s alpha and average interitem correlation. An alpha of ≥0.80 reflects good internal consistency (¹⁶,²⁷). The average interitem correlation is independent of the number of items and sample size, with a value of 0.15–0.50 deemed acceptable (²⁸). Values >0.50 indicate items are repetitive.

Rasch Analysis.

Data were tested for their fit to the Rasch model using the estimation method Joint Maximum Likelihood Estimation in jMetrik, version 4.1.1 (²⁹). The following criteria for the Rasch model were investigated: item statistics (INFIT and OUTFIT mean squares), person and item reliability, rating scale diagnostics, factorial test of residuals, and differential item functioning (nondialysis CKD versus transplant, age, sex, and stage 1–3 versus 4–5). Full details can be found in Supplemental Material 1.

Results

Patient Characteristics

Out of the 1017 participants, 73 did not complete the PAM-13 as part of the survey pack and two were excluded because they had ≥3 missing answers (patient 1: items 1, 2, 3; patient 2: items 1, 12, 13). Overall, this left 942 for analysis. Demographic and clinical characteristic details can be found in Table 1. The average age was 66 (±21) years, 59% (n=554) were male, 95% (n=893) were of White ethnicity. The mean eGFR was 36 (±21) ml/min, and 18% were transplant recipients. Participant characteristics across disease stage and type can be found in Supplemental Material 2.

Table 1. - Characteristics of participants in a multicenter observational study of CKD (DIMENSION-KD) who completed the 13-item Patient Activation Measure

Characteristics	All	Nondialysis CKD	Transplant
	(n=942)	(n=762) a	(n=163) a
Age, yr	66 (±21)	68 (±14)	60 (±12)
Sex male, n (%)	554 (60)	450 (59)	93 (57)
Ethnicity, n (%)
White British	865 (93)	709 (94)	140 (87)
South Asian	22 (2)	14 (2)	7 (4)
Other White	28 (3)	20 (3)	8 (5)
Other	23 (2)	9 (1)	7 (4)
Highest education attained
Tertiary education, n (%)	378 (40)	299 (39)	79 (48)
Secondary education, n (%)	354 (38)	292 (38)	62 (38)
Cardiovascular disease, n (%)	315 (34)	271 (36)	39 (24)
Diabetes type II, n (%)	257 (27)	226 (30)	28 (17)
Liver disease, n (%)	75 (8)	61 (8)	13 (8)
Systolic BP, mm Hg	141 (±45)	139 (±21)	149 (±94)
Diastolic BP, mm Hg	77 (±12)	77 (±12)	79 (±12)
Blood pathology
eGFR, ml/min per 1.73 m2	36 (±21)	32 (±20)	47 (±21)
Hb, g/dl	12.2 (±2.1)	12.2 (±2.1)	12.2 (±2.0)
Albumin, g/dl	4.1 (±0.4)	4.0 (±0.5)	4.1 (±0.4)
CRP, mg/L	10.2 (±15.6)	10.8 (±16.4)	6.4 (±9.8)
Phosphate, mg/dl	4.6 (±21.4)	5.6 (±25.4)	3.1 (±16.4)
Sodium, mEq/L	140 (±5)	140 (±6)	140 (±6)
Iron, ug/dl	71 (±31)	72 (±33)	66 (±18)
Potassium, mEq/L	5.2 (±14.6)	5.3 (±16.7)	4.5 (±0.6)
BMI, kg/m2	28.8 (±7.4)	29.2 (±7.70	26.7 (±5.4)
PAM score	55.1 (±15.1)	55.1 (±14.5)	55.0 (±17.9)
PAM level 1, n (%)	238 (26)	194 (26)	44 (28)
PAM level 2, n (%)	288 (32)	245 (33)	43 (27)
PAM level 3, n (%)	250 (28)	204 (28)	46 (29)
PAM level 4, n (%)	121 (14)	95 (13)	26 (16)

Data presented as mean and SD. Tertiary education includes completion of university or college; secondary education includes completion of high school. Hb, hemoglobin; CRP, C-reactive protein; BMI, body mass index; PAM, Patient Activation Measure.

^aNo CKD group for n=17, although included in main analysis.

Data Quality and Item Analysis

The mean PAM-13 score was 55.1 (±15.1). The distribution of the PAM-13 scores and levels across CKD stage and type can be found in Figure 2 (data in Supplemental Material 3). Two (<1%) patients scored 0; 22 (3%) scored 100. The PAM-13 levels were distributed as Level 1 (n=238, 27%); Level 2 (n=288, 32%); Level 3 (n=250, 28%); and Level 4 (n=121, 14%). Items 1–9 displayed a ceiling effect (i.e., >15% of patients scored the maximum response). No items displayed a floor effect. Item response was high, with a small amount of missing data (all items <1%). Responses to N/A were negligible (all items <5%). All items (apart from 5, 6, 12) displayed moderate to high (left) negative skewness (i.e., toward lower activation) (Figure 3).

Figure 2.: Distribution of the 13-item Patient Activation Measure scores with disease stage (using eGFR thresholds) and type. Ceiling and floor effect defined as responses with >15% for responses “Agree strongly” and “Disagree strongly,” respectively. PAM, Patient Activation Measure.

Figure 3.: Breakdown of responses for each item. Ceiling and floor effect defined as responses with >15% for responses “Agree strongly” and “Disagree strongly,” respectively. NA, Not applicable.

Internal Consistency (Reliability)

Cronbach’s alpha was 0.925 (95% confidence interval, 0.917 to 0.932, SEM, 2.2), and average interitem correlation 0.502. Item-rest correlations (Table 2) ranged from 0.460 to 0.545 and were >0.50 for four items (⁵,⁶,⁸,⁹).

Table 2. - Item descriptives for all patients

Items	Mean (SD)	Skewness (SE)	Missing, n	NA, n (%)	Floor, n (%)	Ceiling, n (%)	Item-rest Correlation
1. When all is said and done, I am the person who is responsible for managing my health condition	3.2 (0.9)	−1.4 (0.1)	1	22 (2)	28 (3)	361 (38) a	0.511 b
2. Taking an active role in my own health care is the most important factor in determining my health and ability to function	3.1 (0.9)	−1.4 (0.1)	4	28 (3)	20 (2)	283 (30) a	0.485
3. I am confident I can take actions that will help prevent or minimize some symptoms or problems associated with my health condition	2.8 (0.9)	−1.2 (0.1)	6	37 (4)	31 (3)	155 (17) a	0.494
4. I know what each of my prescribed medications does	2.9 (0.9)	−1.0 (0.1)	7	21 (2)	30 (3)	222 (24) a	0.500
5. I am confident I can tell when I need to go get medical care and when I can handle a health problem myself	3.0 (0.8)	−0.9 (0.1)	4	8 (1)	23 (3)	230 (25) a	0.545 b
6. I am confident I can tell my health provider the concerns I have even when he or she does not ask	3.0 (0.7)	−0.9 (0.1)	5	9 (1)	19 (2)	234 (25) a	0.544 b
7. I am confident I can follow through on the medical treatment I need to do at home	2.9 (0.9)	−1.3 (0.1)	5	38 (4)	25 (3)	231 (25) a	0.484
8. I understand the nature and causes of my health condition	2.9 (0.8)	−1.0 (0.1)	4	14 (2)	31 (3)	189 (20) a	0.545 b
9. I know the different medical treatment options available for my health condition	2.8 (0.8)	−1.0 (0.1)	4	21 (2)	37 (4)	143 (15)	0.528 b
10. I have been able to maintain the lifestyle changes I have made for my health	2.7 (0.9)	−1.0 (0.1)	4	35 (4)	41 (4)	114 (12)	0.460
11. I know how to prevent further problems with my health condition	2.7 (0.8)	−1.2 (0.1)	9	37 (4)	25 (3)	87 (9)	0.478
12. I am confident I can find a solution when new situations or problems arise with my health condition	2.5 (0.8)	−0.8 (0.1)	5	32 (3)	46 (5)	69 (7)	0.479
13. I am confident I can maintain lifestyles changes, like diet and exercise, even during times of stress	2.6 (0.9)	−1.1 (0.1)	4	34 (4)	46 (5)	88 (9)	0.467

Floor defined as those selected response 1 (Strongly Disagree). Ceiling defined as those selected response 4 (Strongly Agree). For the overall scale, Cronbach’s alpha was 0.925 (95% confidence interval, 0.917 to 0.932, SEM 2.2), and average interitem correlation 0.502. NA, Not applicable.

^a>15% for ceiling effect.

^b>0.500 for item-rest correlation.

Rasch Analysis

Item Statistics.

In our sample, item INFIT/OUTFIT mean square values (indicate how accurately or predictably data fit the model) ranged from 0.82 to 1.36. In total, 11 items had values within the range required for a unidimensional measure. Item 7 (“I am confident I can follow through on the medical treatment I need to do at home”) (1.31) and item 13 (“I am confident I can maintain lifestyles changes, like diet and exercise, even during times of stress”) (1.36) had an OUTFIT value of >1.3. Separation distances (measure of whether items are distinctly separate) of at least 0.15 logits were identified for eight of the 12 separations between items, but not for separations between items 6 and 1; 4 and 7; 9 and 3; 3 and 10; and 13 and 12.

Person and Item Reliability.

The overall Rasch person reliability was 0.99. Rasch item reliability was 0.91. The separation index for persons was 9.48 and for items 3.21.

Rating Scale Diagnostics.

The measures of item difficulty are presented in Table 3. The item location parameter ranged from a low for item 5 (“I am confident I can tell when I need to go get medical care and when I can handle a health problem myself”), which was the easiest item (logits −0.90), to a high for item 12 (“I am confident I can find a solution when new situations or problems arise with my health condition”), which was the most difficult (logits 0.86). The scaling of the PAM-13 items from our analysis should be as follows: 5 (−0.90), 6 (−0.80), 1 (−0.56), 2 (−0.31), 8 (−0.31), 4 (−0.18), 7 (0.01), 9 (0.12), 3 (0.29), 10 (0.52), 11 (0.62), 13 (0.65), and 12 (0.86).

Table 3. - Item calibration for all patients

Items	Difficulty (logits)	SEM	Information-Weighted Fit	Outlier-Sensitive Fit
1. When all is said and done, I am the person who is responsible for managing my health condition	−0.56	0.05	0.98	1.15
2. Taking an active role in my own health care is the most important factor in determining my health and ability to function	−0.31	0.06	0.96	1.22
3. I am confident I can take actions that will help prevent or minimize some symptoms or problems associated with my health condition	0.29	0.05	1.04	1.12
4. I know what each of my prescribed medications does	−0.18	0.06	1.03	1.10
5. I am confident I can tell when I need to go get medical care and when I can handle a health problem myself	–0.90 (least)	0.06	0.85	0.82
6. I am confident I can tell my health provider the concerns I have even when he or she does not ask	−0.80	0.06	0.82	0.79
7. I am confident I can follow through on the medical treatment I need to do at home	0.01	0.05	1.11	1.31
8. I understand the nature and causes of my health condition	−0.31	0.06	0.84	0.82
9. I know the different medical treatment options available for my health condition	0.12	0.06	0.95	0.95
10. I have been able to maintain the lifestyle changes I have made for my health	0.52	0.06	1.16	1.24
11. I know how to prevent further problems with my health condition	0.62	0.06	1.08	1.18
12. I am confident I can find a solution when new situations or problems arise with my health condition	0.86 (most)	0.06	1.08	1.20
13. I am confident I can maintain lifestyles changes, like diet and exercise, even during times of stress	0.65	0.06	1.01	1.36

INFIT is an information-weighted residual of the observed responses from the model expected responses. INFIT mean square error is a quality control fit statistic assessing item dimensionality. OUTFIT mean square error fit statistic is sensitive to unexpected observations made by respondents on items that would be either very easy or very hard for them. It is an outlier-sensitive statistic. INFIT, information-weighted fit; OUTFIT, outlier-sensitive fit.

Structural Validity.

Unidimensionality examined using Principal Component Analysis of Residuals analysis showed the first component explained 30% of the total variance, and the Eigenvalue of the first contrast was >2.0 (2.3). The two items with the strongest positive loadings on the first contrast were item 5 (0.49) and 6 (0.46). The three items with the strongest negative loadings were items 10 (−0.62), 13 (−0.61), and 11 (−0.54).

Differential Item Functioning.

We did not observe large differences (defined as class CC) in differential item functioning for CKD stage (1–3 versus 4–5) or sex. The differential item functioning test for disease type (nondialysis CKD versus transplant) showed large differences for item 3 (chi square, 3.21, P=0.07, CC-), that is, the item is more difficult for the nondialysis CKD group than for the transplant group. Dividing the sample into two age groups (on the basis of median age), ≤69 years versus >69 years, a statistically significant difference was observed for item 12 (chi square, 7.32, P=0.01), indicating this item is easier for the ≤69 years group. Full differential item functioning analysis can be found in Supplemental Material 4.

Discussion

The aim of this study was to assess the psychometric properties of the PAM-13 in people with CKD. Overall, the PAM-13 performed reasonably well in some areas, but not all. The items had good fit statistics and largely conformed to unidimensionality. The person and item reliability indices were high, suggesting orderings were replicable. However, item scaling revealed considerable ceiling effects. We did not observe the presence of differential item functioning for stage or sex, although differential item functioning was present for some items between disease type and age. In the absence of a kidney-specific instrument, the PAM-13 remains the best measure to assess patient activation, although caution is warranted for several items.

Despite the PAM-13 being used in a growing number of studies, the relevance to individuals with CKD is unknown. In our data, the small amount of N/A responses (≤5%) suggests the items were appropriate. The first nine items had a ceiling effect, suggesting the four-point Likert scale is not relevant or sensitive for this population. High ceiling effects have previously been observed, particularly for “easier” Level 1 items (¹⁶), and may reduce the ability to detect changes over time (e.g., in interventional studies).

Despite the ceiling effects, our findings suggest the response structure (i.e., properly ordered, well-defined, and mutually exclusive categories [³⁰,³¹]) was acceptable and advanced monotonically. Our person reliability (>0.9) indicated an ability to separate the scale into three or four levels/categories, unlike others (¹¹,¹⁶), who could only distinguish between two and three levels/categories. We found each item could be regarded as part of one dimension (INFIT/OUTFIT) performing as a Guttmann sum-scale. However, item 7 (“I am confident I can follow through on the medical treatment I need to do at home”) and item 13 (“I am confident I can maintain lifestyles changes, like diet and exercise, even during times of stress”) had an OUTFIT value >1.3, indicating the items are either poorly constructed, ambiguously defined, or do not define the same construct as the rest of the items. Further evaluation is needed to determine how patients interpret these items, although it may be that the role of self-management (item 7) and lifestyle (item 13) are poorly understood in CKD. Although high item reliability may indicate the sample is big enough to precisely locate the items on the latent variable (¹¹,¹⁶), these estimates also indicate the scale is appropriate on an individual basis to diagnose activation and individualize plans for future health care (³²).

We found good internal consistency, although high interitem correlation suggests repetition between items, in particular, items 5, 6, 8, and 9. In these items, there seemed to be no additional information when answering the next item in the scale (i.e., no separation). Separation between item locations should be >0.15 logits, and a value <0.15 logits indicates redundancy (³³). On the basis of our findings, three to four items could be omitted from the scale as a simple sum-scale. However, the overall INFIT/OUTFIT values for the 13 items were within the recommended interval (0.5–1.5), thus fulfilling the criterion for item goodness-of-fit. Findings from other European countries have also demonstrated acceptable INFIT/OUTFIT values (¹⁴,¹⁶).

Similar to Maindal et al. (¹⁶), who found the item order did not represent consecutive item difficulty for their population, we identified issues related to the PAM-13 scaling (i.e., item order). Our population, like others (¹⁶), found some questions easier to answer. As others did (¹¹,²⁰), we found item 3 (“I am confident that I can help prevent or reduce problems associated with my health”) required a higher level of activation (i.e., was more difficult to achieve) than subsequent items. This may reflect the somewhat progressive immedicable nature of CKD that patients feel they cannot alleviate. We found, like others (¹¹,²⁰), item 4 (“I know what each of my prescribed medications does”) was easier than item 2 (“Taking an active role in my own health care is the single most important thing affecting my health”). This suggests that patients with CKD feel somewhat passive in their health management. Consequently, this may affect the classification between Levels 1 and 2 and create uncertainty about individuals’ abilities to participate and act. Difficulties in classification have previously been reported, where 40% patients were incorrectly rated as Level 1 rather than Level 2 (²⁰). Given that guidelines for using the PAM-13 to tailor care recommend focusing of different aspects of activation, it may be more appropriate to tailor interventions to the combined needs of both Levels 1 and 2. We did not find item 7 (“I am confident that I can follow through on medical treatments I need to do at home”) required a lower level of activation than earlier items, unlike others, who found it was ranked first (²³) and fourth (²⁰) easiest. This is perhaps unsurprising, given medication adherence rates are lower for those with chronic diseases compared with those with neurologic conditions (³⁴). Consideration of differences in self-management tasks between diseases and populations may be an important factor when measuring activation (²⁰).

Consistent with others (¹,¹¹), we found the last five items were the most difficult, although we also found item 3 was difficult. Inconsistencies between scaling have previously been reported (¹¹,^1415–16,²⁰). Patient responses to items have been shown to be affected by response style, cultural influence, meaning/interpretation of the items, prognosis and trajectory of the condition, and the health care system (¹²,¹⁴,¹⁶,²⁰,³⁵). Because patient activation may not only be influenced by knowledge, skills, and confidence, but also other factors (³⁶), a comprehensive understanding of the patients’ beliefs, health care professional practices, and health care systems is needed (¹¹), and we need to ensure the items and response options are relevant and understood as intended (³⁷).

When testing for differential item functioning, we did not observe large differences between disease stage or sex. No differences between sexes have been reported previously (¹¹). It may be assumed that self-management requirements change as disease progresses, although given the relatively nonspecific statements, patients may be able to appropriately adapt perceptions of their current health status with each item. Our findings showed the presence of differential item functioning for item 3 (“I am confident I can take actions that will help prevent or minimize some symptoms or problems associated with my health condition”) between disease groups, with this item more difficult in the nondialysis CKD group. Although the PAM-13 scores were identical between groups, this suggests that transplant recipients are more confident in taking actions to manage their health. We saw a large differential item functioning for item 12 (“I am confident I can find a solution when new situations or problems arise with my health condition”) between age groups, suggesting older patients may lack confidence in dealing with unfamiliar or new health problems.

In the absence of a disease-specific instrument, with reliable and relevant items, our results support the use of the PAM-13 to assess activation levels, and as a screening tool for tailoring self-management interventions or a quality indicator for delivery of care (⁹). However, consideration should be used for several items, and responses to individual items may help better guide specific clinical management than using the PAM levels generated alone. Self-management tasks required for CKD, particularly for more advanced CKD, may not be adequately captured using the PAM-13, and thus additional CKD-specific self-management questions may be required. Using the PAM-13 could allow for a more customized approach to tailoring interventions and preparing appropriate education. Tailored care after activation assessment has resulted in improved clinical indicators and decreased health care utilization in patients with chronic disease (³⁸). In addition, tailoring care to activation levels may facilitate more productive health care interactions by empowering patients (³⁹). For future scale revision, we suggest two areas for consideration: (1) collapse categories 1 and 2 for all items to improve parameter estimation, and (2) add some high-end items to the scale to cover the upper end of the trait. If the originally proposed model (¹) is to be of significance to patients with CKD, items could be reordered to represent consecutive item difficulty. Modifying the PAM-13 to be condition specific has important implications that could result in the loss of the ability to compare populations and the relative effect of different medical conditions and/or treatments. Any modification of the PAM-13 could violate any licensing agreement, although development of a disease-specific PAM-13 could be supported by the licensee.

There is growing interest in the utility of the PAM-13, particularly in the United Kingdom where the PAM-13 is being appraised as a tool used to evaluate care for chronic conditions in the NHS, including NHS England’s self-care program, by measuring and responding to peoples’ activation levels (⁵,⁹). The PAM-13 is presently the only validated, evidence-based tailoring tool to support services in building individuals’ skills, knowledge, and confidence to manage their health. However, the utility of the PAM-13 must be considered, because the PAM-13 assesses an individual’s perceived ability to engage in self-management, not the individual's actual ability. Although Zimbudzi et al. (⁴⁰) found the PAM-13 was associated with self-care activities in patients with diabetes and CKD, these were not objectively assessed. Further research is needed to establish the association of the PAM-13 with self-management tasks.

In England, the license cost associated with the PAM-13 is funded by NHS England and NHS Improvement as part of a national agreement. In the United States, PAM has been included in the Kidney Care Quality Alliance framework and in Kidney Care Choices payment model of the Centers for Medicare and Medicaid Services (Supplemental References).

The PAM-13 was delivered as part of a larger survey and the number of missing values, although low, may have been reduced if the PAM-13 was delivered on its own because the larger survey may have fatigued respondents. Because the PAM-13 is designed to be applicable across multiple conditions, patients may have felt some items were irrelevant to their health. Individuals with comorbidities may have found it difficult to separate their responses for CKD from other related conditions (i.e., diabetes, hypertension), although managing multimorbidity is a frequent occurrence. The findings may have some generalizable limitations; with surveys returned from different centers across England as part of a research study, our sample was predominantly White patients (>90%), and thus the findings may not be generalizable to the general kidney population, minority ethnic groups, or countries with dissimilar health care systems to the United Kingdom.

In summary, our findings show the PAM-13 has high reliability and forms a unidimensional, Guttman-like scale. Although the PAM-13 appears to be a suitably reliable and valid instrument for assessing patient activation in CKD, the high ceiling effect may be a problem when measuring change over time. High interitem correlation and interitem separation suggests repetition between items and potential for redundancy. The PAM-13 has promising psychometric properties, indicating it can be used; however, further validation in other chronic disease populations is advisable. Future research should focus on interventions to increase activation to improve health care outcomes.

Disclosures

A. Smith, T. Wilkinson, and C. Lightfoot report employment with University of Leicester. J. Palmer reports employment with University Hospitals of Leicester, University of Leicester. T. Wilkinson reports consultancy agreements with Baxter Healthcare Ltd. and reports receiving honoraria from Baxter Healthcare Ltd. K.E. Memory reports support from a Kidney Research UK Intercalating Student Bursary.

Funding

This research was funded by the National Institutes of Health Research Applied Research Collaboration East Midland and the Stoneygate Trust and supported by the National Institute for Health Research Leicester Biomedical Research Centre (BRC). K.E. Memory is supported by a Kidney Research UK Intercalating Student Bursary.