Post-stroke Quality of Life Index: A quality of life tool for stroke survivors from Sri Lanka

Stroke is a rapid vascular event with symptoms lasting 24 h or longer [1]. Stroke is regarded as a major cause of disability anywhere in the world [2]. Its burden has been relatively worsening in low- and middle-income-countries (LMICs) compared to high-income countries (HICs) in relation to its incidence as well as social and economic impacts [3]. As an example when a decrease of 42% of stroke incidence is observed in HICs, an increase over 100% is observed in LMICs over the last four decades [3]. Sri Lanka was recently upgraded as a Upper-Middle Income Country after being a LMIC with comparatively satisfactory health parameters [4, 5]. Prevalence of stroke in urban Sri Lanka was estimated to be 10.4 per 1000 (95% CI = 6.3 to 14.5) with a male to female ratio of 2:1 [6]. Relatively higher incidence of stroke among the young also been noted here [7]. With the demographic and the epidemiological transitions, as in many other LMICs, stroke burden is expected to rise further in future.

The World Health Organization (WHO) defines Quality of Life (QOL) as “the individual’s perception of his/her position in life in the context of the culture and value systems in which they live and in relation to their goals” [8]. It is a composite concept affected by many facets [8]. In a patient with a disease condition, the QOL may reflect the success of management of that particular disease, as the patient perceives [9]. Disabilities imposed by stroke would directly or indirectly influence the physical as well as psychological QOL components as perceived by the patients. Hence the QOL of stroke survivors is an extremely important factor, for research, to assess progress and to target services for stroke survivors [10‐12].

Since QOL is a patient-reported outcome, there is no agreement on exactly which domains are to be captured within its scope [13]. Hence, though there are many available QOL tools, there is much variability among their domains [14, 15]. Environmental influences (such as user-friendliness and safety of household items) and economic influences (such as financial security), do affect the “self-perceived position in life”, especially in disability-related conditions like stroke. Many QOL tools do not capture these, even though these are proved to impose an impact on overall QOL [16‐22]. It has been encouraged to design QOL instruments specially including domains like economic costs and burden to the family members [23]. Furthermore, epidemiological characteristics of stroke are prone to vary even within LMICs [24, 25]. The living contexts (i.e. standards of household environment and items) and the expectations of the people (i.e. due to the differences in salary scales and in social insurance systems) are different between the HICs and LMICs. Stroke-related literature is scarce in relation to LMICs. Hence most of the evidence on stroke rehabilitation which come from Western settings which might not be applicable for lower-middle income settings [26‐30]. The development of new QOL tools for stroke survivors would enable more context-related QOL measurements [31].

Out of the many health-related QOL tools available, Short-Form-36 (SF-36) and European Quality of Life 5 dimensional (EQ-5D) tool are two commonly used generic tools [9, 32‐34]. SF-36 includes 08 scales and has been used to assess QOL in stroke survivors following hospital discharge [35‐39]. It has been validated for several disease conditions including stroke within Sri Lanka [37, 40]. EQ-5D tool consists of an EQ-5D descriptive system as well as an EQ-Visual-Analogue-Scale (VAS). The former system include five dimensions and the latter records the respondent’s self-rated health on a VAS [41]. It too has been validated within Sri Lanka [42, 43]. SF-36 covers a period of 28 days previous to the date of data collection and the EQ-5D captures the QOL related to the time of data collection.

Validity, reliability and responsiveness are three properties that must be evaluated in relation to a newly developed measurement tool [44, 45]. Validity measures whether the tool “actually measures what it is expected to measure” [46]. In the absence of a gold standard test for QOL, the validity based on data is assessed by the construct validity [47]. Reliability refers to a measure of inherent amount of error of any measurement [48]. It includes the assessment of homogeneity of each of the items measured by Cronbach Alpha and reproducibility (measured with test-retest method etc.) [48, 49]. Responsiveness is defined as the ability to detect a clinically important change [50, 51].

Sri Lanka is trying to expand stroke-related services in general including that of acute care, rehabilitation and community-based care [52]. Availability of context-specific QOL tools would greatly help in assessing the success of management of stroke survivors in this context. The aim of the study was to develop and validate a comprehensive quality of life tool for stroke survivors at post-stroke 1-month in Sri Lanka.

The domains of QOL that have been included in the tools found in literature search have been listed in Supplementary file 2 [10, 53, 64‐68].

After combining the related items, the initial list included 36 items. The mean and median cumulative scores given by the expert panel through the modified-Delphi technique for the initial item list was 22 and 28 respectively. The upper margin of the first quartile was 7. The number of items which scored more than the upper first quartile was 21. Figure 2 shows the distribution of the cumulative scores given by the experts. No outliers were detected for the scores on clarity and wording.

The factor structure analysis of the 21 items is mentioned in Table 1.

Considering the factor loadings and the correlations of the factors, items were assigned into the respective factors. The complex variables were retained in the best factor as suggested by the Cronbach alpha values with “items removed”. The Table 2 summarizes the correspondence of the retained items to the individual domains.

The developed tool was named as “Post-stroke Quality Of Life Index for strokes survivors” (PQOLI) (Table 3).

The measures of location and measures of dispersion were obtained as shown in Table 4. All domain scores have a potential range from 0 to 100.

Table 5 shows the spearman-correlation of the domain scores with the scores of SF-36 and the EQ-5D tools.

All domains showed statistically significant positive correlations with scores of SF-36 and EQ-5D-3 L. The EQ-5D-3 L scores had relatively higher strength of associations for the Domains I and IV. In general, for the domain scores with more similar constructs of the SF-36, the associations were with relatively higher strengths.

The ability to discriminate between the two ends of the severity spectrum of a disease entity is mentioned in Table 6. It shows that the QOL was significantly higher in the ambulatory group than the institutionalized group.

In the study sample, 27%(n = 49) had to apply for a loan, 30%(n = 54) had to sell a property, 67%(n = 120) had a potential income loss and 67% (n = 120) had to restrict other expenses, due to the illness. Approximately 73% (n = 133) had a financial burden. Higher values for the financial-domain-score were obtained by those who did not experience a financial burden. The median (IQR) financial-domain-scores of PQOLI for the group with financial burden was 35.00 (25.0–50.0). The relevant scores of the group without financial burden was 40.00 (35.0–55.0). There was a statistically significant difference between these two groups (p = 0.012).

Cronbach alpha level for the domains I,II,II and IV were respectively 0.906, 0.880, 0.803 and 0.682.

The PQOLI was developed upon context specific evidence of the study setting. It includes 21 items under four domains of physical/social, pain/emotional, financial stability and environmental domains. Hence it addresses many of the deficiencies of utilizing current QOL measures within the resource-constrained settings [26‐29, 31].

Many of the recommended methodological aspects in the development of QOL tools were adhered in the development of the PQOLI [44]. As recommended in global literature, patients too were involved in several components of the study including the phase of item generation [33, 69]. The factorability assessments done by KMO measure, Barlett test, item communalities and anti-correlation images were satisfactory [57, 70, 71]. The KMO measure was more than 0.8 and hence can be classified as “meritorious” [57]. The significant Barlett test suggests that the “correlation matrix was significantly different from the identity matrix and therefore factorable” [57].

Exploratory Factor Analysis is better suited when the domain structure is not previously known as in this study [70]. Principal Component Analysis assumes the continuous nature of the variables and it was ensured by the presence of six response categories for each item [71]. It was performed following that with the intention of having a smaller number of variables that explain the most variation in the original set. Even though the orthogonal varimax rotation is commonly done, as in the present study, oblique techniques have been recommended in some literature [69, 72]. A minimal eigenvalue of 1 is traditionally used in defining the factors as in the present study [61]. All the items had a loading value more than the traditional cut-off of 0.32 [69]. The complex factors with cross-loadings were dealt with internal consistency analysis [69, 72, 73] The Cronbach’s alpha values with “items removed” suggest that the allocation of cross loadings were accurate [74].

PQOLI is a disease-specific QOL tool in contrast to generic tools [75]. Disease-specific tools are considered better than generic tools for being more sensitive in picking the changes of stroke patients [33]. The present tool was developed to cover a recall period of seven days in contrast to the four-weeks recall period of SF-36 and momentary capturing of QOL by EQ-5D [35, 41]. The validity of PQOLI was assessed at one month from the end of acute-management of the stroke. In literature post-stroke QOL has been measured from one month up to five years [76]. The seven-day recall period and utilization at one-month were decided in considering the rehabilitation procedure of the study settings and by considering the quality of patient-responses in the pre-testing period.

Following the factor analysis, financial and environmental components were identified as two domains in the PQOLI. This is an example of the recent global recommendation that domain structures for QOL tools for LMICs should be context specific [31]. Furthermore this proves the necessity of incorporating economic and environmental domains in QOL measurements [16‐22]. In many LMICs like Sri Lanka, social-insurance systems are not found. The services not free-of-charge, may have to be achieved by out-of-pocket expenditure creating a potential financial burden. These may be the reasons for getting a financial-independence domain following factor analysis. Domain II proves that the stroke-survivor-friendliness of the lining environment and the accessibility to healthcare too are important in determining the quality of life.

The pattern of correlations, provides evidence that the a-priori hypotheses no.1 to no.4 are fulfilled. This reflects that the PQOLI was based on a reflective model in addition to the assessment with the “thought test” as recommended in the COSMIN checklist [44]. All the items can be assumed to be changed when the underlying construct changes [44]. In evaluating the construct validity by the correlations, significance, direction and effect size are generally considered [44]. However due to the “complexities of the constructs”, the strength of associations cannot be expected to be interpreted in the exact way as in bivariate correlations. Since then, for the a priori-hypotheses 1 to 3, the significance and the directions were included while the explorations of strengths were focused in hypothesis no.4 [77, 78].

The hypothesis no.4 was fulfilled in relation to domain I, but was not to a great extent for domain IV. The domain structure in the PQOLI and the SF-36 are not the same. As an example, the pain domain which is a separate entity in the SF-36, has been included in the domain IV of the PQOLI together with items related to emotional-health. This difference of the item-structure might be one reason for PQOLI domain IV not having a greater strength with the mental-health, vitality and role-limitation-emotional domains of SF-36. However, the strength of associations were still of “acceptable and of medium strength” [77‐79].

As expected the domain scores of PQOLI were significantly higher for the ambulatory group than the hospitalized group. This highlights its ability to discriminate between the severities of a spectrum. Institutionalized patients are with a higher severity of illness than the patients in the ambulatory group whom have been discharged from the in-ward care. This fulfills the a-prori hypothesis no.5. Hypothesis no.6 tested the ability of the Domain III to discriminate between groups with differences in the level of financial-independence. A lower financial-independence in the group with “financial-burden” would have lowered their subjective perception of the position of living, thus lowering the QOL. This has been accurately captured by the Domain III of PQOLI. In a summary, five of the six a-priori hypotheses have been completely fulfilled and one has been partially fulfilled. Since the alpha values for domains I,II and III were more than 0.8 and for domain IV being closer to 0.7, it can be classified as demonstrating satisfactory internal consistency [44, 70, 80].

Several limitations can be mentioned in this process. One is that the PQOLI provides only numerical outcome-scores for the domains without providing a categorical outcome as “satisfactory” and “not-satisfactory”. This is acceptable as establishing a cut-off needs research in larger scale and it would be a future extension of this. Secondly the PQOLI provides 4 domain scores rather than an amalgamated score. Such a score would need weighing of the domains and that too would be done as a future extension of this research which would be done in a larger scale. The tool underwent exploratory factor analysis and it was not followed by confirmatory factor analysis in another study sample. Hence it is advisable to conduct confirmatory-factor analysis when this is used in another setting.

The reproducibility was not assessed for PQOLI and only the internal consistency reliability was assessed [80‐82]. Similarly the responsiveness of the instrument was also not assessed [50, 51]. These reflect further directions on the future research o PQOLI. There were limited descriptive data on the characteristics of the study sample. Though this was done to minimize the time of data collection, it is another limitation of the study. However, in order to minimize any selection bias, participants with a history of physical or mental conditions that would affect the quality of life were excluded. Furthermore, the study settings were with free healthcare and were without any restriction of access [83].

PQOLI included 21 items which are categorized under 04 domains in its development. There is first evidence for sufficient construct validity of the PQOLI as a valid QOL tool for measuring the QOL of stroke survivors with satisfactory internal consistency reliability, when assessed using already validated QOL tools and with the “known group comparison method”. Five out of the six a-prori hypotheses were completely fulfilled in testing it for validity. Its internal consistency reliability was reflected to be satisfactory. PQOLI can be used for the assessment of QOL after 1 month from the end of acute-phase of management of the stroke survivors, following further explorations.