Assessment of the psychometric properties and refinement of the Health and Self-Management in Diabetes Questionnaire (HASMID)

Diabetes is a complex condition in which the long-term consequences of the disease are largely determined by the ability of the individual to self-manage dietary habits, physical activity and various medications. In terms of glucose control, people with diabetes (PwD) need to avoid chronic hyperglycaemia (high glucose levels in the bloodstream) to evade microvascular diabetes-related complications which include eye damage (retinopathy), kidney damage (nephropathy) and neuropathy (nerve damage), which if severe can lead to blindness, dialysis and leg amputations, respectively. Macrovascular complications such as heart attacks and strokes are also increased as a result of chronic hyperglycaemia. Some medications can cause hypoglycaemia (low blood glucose levels in the bloodstream), which if mild may cause for example sweating, slurred speech and tingling, but if more severe can cause confusion, loss of consciousness, seizures and occasionally may be fatal. Many people with type 2 diabetes (T2DM) control their glucose through adapting their diet or taking tablets that do not cause hypoglycaemia. For some, with longer duration of T2DM, medications may be required which if not given at just the right dose may cause hypoglycaemia (including insulin). For people with type 1 diabetes (T1DM), insulin treatment is the only option, and doses need to be adjusted according to the amount of carbohydrate in each meal/drink (as this is the primary source of glucose), the current glucose level, the intensity of physical activity (prior and post the injection) and whether or not there has been any recent alcohol consumption. As there are several factors to take into consideration multiple times each day it is not unsurprising that avoiding hyper- and hypo-glycaemia is more challenging in T1DM than in T2DM. In all forms of the disease (T1DM accounts for approximately 5%, T2DM approximately 90%, other approximately 5%), additional aspects of diabetes care may include medication for raised blood pressure and/or raised cholesterol, as well as therapy for the diabetes-related complications. Hence, the ability of the patient to self-manage their condition is of paramount importance to help them live as healthy a life as possible. Some PwD live a normal life-span with minimal complications of diabetes, whereas others suffer devastating complications and die early [1]. Therefore, educational interventions which aim to improve self-management skills such as DESMOND for T2DM, and DAFNE [2] for T1DM, are seen as the cornerstone of good diabetes care [3].

In a previous study, we confirmed that self-management plays a significant role in the life of PwD [4]. During the project we developed a patient reported outcome (PRO) measure that was able to measure the quality of life impact of self-managing the condition capturing both health and treatment experience. The Health And Self-Management In Diabetes (HASMID^v1) questionnaire consists of 8 attributes, 4 about quality of life, and 4 about self-management [4]. However, psychometric analyses around the performance of the questionnaire have not been previously undertaken. The validation of all PROs is important to demonstrate that they measure what they are intended to measure. The psychometric performance of PROs should be assessed and reported so that users can be assured that the instrument is appropriate to use. The overall aim of this study was to rigorously examine the psychometric properties of the HASMID^v1 questionnaire. The study had two phases: Phase 1 identified any items of the HASMID^v1 questionnaire that may require rewording and deriving alternative and any additional items; Phase 2 conducted an assessment of the measurement properties of the HASMID^v1 questionnaire in a large observational survey of people with diabetes. The aim was to examine floor/ceiling effects, known-group differences and internal consistency.

In the development of any PRO it is important to fully evaluate its performance prior to mainstream usage. The present study has allowed us to further improve on the face and content validity of the HASMID instrument. The original questionnaire was developed using a mixed methods approach, with items generated from interviews with PwD and an existing PRO measure [4]. Here we have been able to re-examine the content of the HASMID questionnaire through Patient and Public Involvement consultation and two focus group cognitive debriefing exercises. The alternatively phrased items were then subjected to psychometric evaluation alongside the existing HASMID items.

A limitation of the study is the method of data collection. Every effort was made to ensure that participants did not complete the online survey twice (to increase their chances in the prize draw). We removed responses within the same time period that provided a duplicate email address. However, it was not possible to identify people who may have completed the survey twice in the same period who used different email addresses. We did note that the sociodemographic and health profiles of PwD differed depending upon whether they completed the online or postal survey. The implications of this are discussed further in Rowen et al. (2019) [16]. Participants were self-selecting, and this too could be considered a limitation.

The development, refinement and evaluation of PRO instruments can be driven by different theoretical approaches: Classical Test Theory (CTT) Item Response Theory (IRT) and Rasch Measurement Theory (RMT) [17]. Here we applied CTT and RMT to assess the performance of the alternatively worded items. Both assessments were used, in conjunction with clinical opinion to inform the final selection of items. This approach has been used by others in PRO development [18, 19]. There are benefits to considering alternative sources of evidence. One of the main indicators for the inclusion of Rasch analysis in this study was to identify whether items demonstrated DIF. There is logic to considering eliminating items based upon differing responses driven by gender and age. However, in the context of self-management of diabetes, age and gender may well be factors that drive individuals’ responses. For example, being tied to mealtimes may be a negative issue for the younger generation, whereas older people may already follow a more structured routine. The benefit of including psychometric analyses, Rasch analyses and clinical opinion allows a consideration of the relevance of the inclusion/exclusion of each item. Adopting only one approach may result in an instrument that is not relevant to the target population or provide information that is not useful in trials, service evaluations, or routine clinical care.

The alternative items to the original HASMID items tested in this study were felt to be exploring the same concept, but with slightly different phrasing. For three (of the five) alternatively phrased items there were insufficient evidence to suggest an amendment to the wording of the original HASMID item. Two of the alternative items have now been incorporated into the HASMID-10 questionnaire – Irritable and Social Activities. Both items performed well psychometrically, were deemed of clinical importance and captured different severity of underlying health to the original items.

One of the limitations of the study was that respondents self-reported information about their diabetes including diabetes type, duration of diabetes, HbA1c status, treatment and diabetes-related complications. Our HbA1c data indicates that respondents often did not know their HbA1c status, meaning that this cannot be reliably used to assess how HASMID performs across different levels of HbA1c, and we have no objective measure of severity by which to examine the items.

A potential limitation to the study is the applicability and performance of items across different ethnic groups. Within the psychometric survey information on ethnicity was not collected. This was a purposeful omission. It was outside the scope of the study to assess the cultural validity of the overall questionnaire and all potential items. To do so would require further qualitative work across different ethnic groups, specifically on cognitive debriefing to ensure items are relevant, and to identify any potential new items for consideration. Therefore, it was felt that to collect data on the ethnicity of survey respondents would not be relevant. Any difference in HASMID scores across groups could not be validated at this stage. Further research is required to assess the cross-cultural validity of the HASMID-10 amongst different populations.

This cross-sectional validation study has examined the psychometric properties of the original HASMID items and tested potential items for inclusion using a large dataset. Rasch analysis was undertaken and considered alongside conventional psychometric performance and clinical opinion. The analyses found the items to have good psychometric performance, with discriminative validity to be able to discriminate across type of diabetes. However further assessment of psychometric performance is recommended by administering the measure alongside a clinical intervention. The final HASMID questionnaire now consists of ten items, the HASMID-10. The additional two items may provide further insight into how PwD are self-managing their condition, by providing further detail into how emotions and daily activities are affected.