Statement of principal findings
The aim of this paper was to report the development and preliminary validation of a new measure of empowerment for patients with long-term conditions in primary care using data from two empirical studies – a qualitative interview study and a cross-sectional quantitative survey.
Being ‘empowered’ to manage a long-term condition was demonstrated in the qualitative data by five dimensions: identity; knowledge and understanding; personal control; decision-making; and enabling other patients with long-term conditions. Questionnaire items were developed in relation to each outcome and tested in a cross-sectional study.
The 5 dimensions identified through conceptual analysis and qualitative work in study 1 were not confirmed empirically in study 2, and a three factor solution was considered to have the simplest structure but this solution was clearly suboptimal and in need of future testing. We applied preliminary labels of ‘positive attitude and sense of control’ and ‘knowledge and confidence in decision making’. The third factor was a complex mixture of items and no clear label was possible.
Given the unclear nature of the three factor solution, we suggest future use of the new measure be restricted to the total empowerment scale, pending future work on the scale structure.
Strengths and weaknesses of the study
The scale was developed based on a literature review (not reported here) and in-depth qualitative work, thereby enhancing the content validity of the measure. The preliminary pilot testing involved interviews with patients and input from a patient and public involvement group to assess acceptability. However, these sources were known to the author and there may have been a problem of social desirability in their response about item content. We did not use the popular technique of cognitive interviewing to test the items, which would have provided a more rigorous test of comprehension .
Test retest reliability of the new scale has yet to be evaluated. The assessment of validity was based on hypotheses regarding associations between the new measure and various comparators. The assessment of validity presented here was based on cross-sectional data and the ability of the scale to predict outcomes longitudinally remains to be demonstrated. A significant number of analyses were conducted, raising issues of multiple hypotheses testing, which might have been avoided through adjustment of the significance level used. However, given the exploratory nature of the work, we wanted to avoid missing any potentially important relationships early into the development process.
All participants were recruited from three general practices and within an inner-city area of North West of England. Sampling was restricted to responders who opted-in to the interview study and on completion of a postal survey. As a result, the respondents in both studies 1 and 2 were relatively homogenous in terms of demographic characteristics (older age, white, retired due to ill health, and located in relatively deprived areas). Ethical requirements meant that respondents had to opt-in to both studies, and such participants may demonstrate certain characteristics, whereas a more diverse sample may have given different results. The potential for bias may be especially high in terms of the limited range of deprivation in the sample, as patients from less deprived backgrounds may have markedly different views.
The response rate was relatively low in study 2, although largely in line with current surveys in this population locally  and nationally . However, it should be noted that the study was not designed to assess prevalence, and that response bias will likely have more limited effects on assessment of relationships between variables (although it may lead to restriction in range if certain types of patients do not respond). It was not possible to explore characteristics of non-respondents as current ethical guidelines do not allow data recording on patients who do not consent to participate.
Interpretation of the results
As outlined previously, the existence of five distinct dimensions of empowerment suggested by the qualitative work were not supported by the quantitative findings in study 2. The factor analyses suggested that responses to items concerning ‘identity’ and ‘control’ were related, as were issues of ‘knowledge and understanding’ and ‘decision making’, and these relationships make conceptual sense.
The meaning of the third factor was very unclear, but the fact that items related to ‘enabling others’ only loaded on this factor might suggest that further tests using only these items would be useful, to see if the validity of this factor was supported. The factor was derived from the patient interviews, and is interesting as it relates to current self-management initiatives in the NHS, such as the Expert Patients Programme , and the Health Trainers initiative , which both rely on non-professionals to teach and empower patients who may be coping less well with their long-term conditions . We would suggest that further psychometric work on this scale is indicated as it has potential wider utility.
Given the ambiguities over the factor solution, we restricted our initial analyses of construct validity to the total empowerment score. In terms of construct validity, the majority of hypotheses were supported from the regression results. Seeing a preferred GP, being educationally qualified and general health were strong predictors of increases in total empowerment. Hypotheses that were not confirmed included the importance of continuity of care and GP confidence and total empowerment.
It is noteworthy that seeing a preferred GP emerged as a key predictor in the analysis, but continuity with the doctor and GP confidence were found to be weak predictors of empowerment. It should be noted that nearly half (49.8%) of responders had not seen their GP/practice nurse for at least 7 to 9 months, which may have introduced a bias in responses on empowerment items that related to GP variables. Only two aspects of continuity were included in this study: seeing the same GP and seeing a preferred GP. It is possible that other measures of continuity may have given different results. For instance, twelve continuity measures have been developed to measure various types of continuity , each emphasizing different elements of the patient-practitioner relationship, such as density of visits and subjective perception of visit . It may also reflect the fact that nurses provide the bulk of long-term condition care in the United Kingdom.
Questionnaire development and validation is an ongoing process, and there are a number of potential developments of the proposed new empowerment measure.
Future testing of the measure may also benefit from confirmatory factor analysis to more rigorously test hypotheses about scale structure, given that ambiguities remain concerning the validity of the three factor solution and the concern over scale length.
We are aware routine use of the new measure is likely to be enhanced by reducing the number of items. Following future large-scale validation of the measure, a short-form version should be developed and tested to lower response burden and increase the possibility of routine use of the measure. The development and testing of the short-version should follow state of the art methodology for shortening composite measurement scales .
If additional supportive evidence is generated, the scale might usefully be used in longitudinal studies or randomised trials of self-care interventions, to see whether it is sensitive to the effects of interventions designed to improve empowerment, and whether such changes are subsequently associated with changes in self-care behaviours, health outcomes and quality of life. Interventions that might be expected to lead to increased empowerment might include: GP communication skills training ; patient decision aid interventions ; and self-management support, such as the chronic disease self-management programme .
Future research might also explore the relationship between measures of illness perceptions, such as the Illness Perception Questionnaire , which is designed to capture representations of specific illnesses, rather than generic feelings of empowerment. Exploring relationships between the measures may be fruitful, as it is possible that certain types of illness representations (such as those around controllability) may be predictive of levels of empowerment.
The new measure has similarities with the Patient Activation Measure  and a formal comparison might highlight advantages and disadvantages of each. Both measures have different psychometric properties and underlying scale structure. The Patient Activation Measure  has stronger psychometric properties than the current measure and was developed from the Rasch Rating Scale Model , an alternative statistical method to factor analysis, used to test scale structure. As a result, the elements of knowledge, belief, and skill that constitute activation have a hierarchical order; thus what is needed to increase activation depends on where the person is on the activation continuum.
There are many other factors not measured in the current study that could be used to assess construct validity. For instance, a quality of life measure or other measures of psychological functioning may have been useful. It was also evident from the qualitative data (not presented here), that health literacy was a key issue in this group of patients, and this could have been examined by administering a health literacy measure . However, such scales can be difficult to use in the context of a postal survey.