Potential use of telephone-based survey for non-communicable disease surveillance in Sri Lanka

Non-communicable diseases (NCDs) are a group of illness that are not due to infectious process and hence are not transmittable from a patient to a healthy person [1]. NCDs account for most of the deaths worldwide, killing more people each year than all other causes combined [2]. Among NCDs, cardiovascular diseases account for most deaths, or 17.5 million people annually, followed by cancers (8.2 million), respiratory diseases (4 million), and diabetes (1.5 million) [3]. Contrary to the popular opinion, people living in low and middle-income countries are disproportionately affected by NCDs and it is estimated that two of every three deaths due to NCDs occur in these countries [4].

Among low and middle income countries, South Asia is one of the worst affected region with NCDs in the world [5]. The prevalence and the adverse health outcomes of NCDs have risen more rapidly in South Asia than in any other geographical region in the world [6]. Even though, demographic dimension of Sri Lanka is different to the rest of the South Asian countries, factors which influence the burden of NCDs are more or less similar [7, 8]. Like the rest of the other nations, there is ongoing rapid urbanization, which has led to unhealthy eating habits, physical inactivity, high rates of smoking and tobacco use in Sri Lanka [7‐9]. Urban population enjoys comparatively better income, but, has more westernized lifestyle with physical inactivity and unhealthy eating habits [10]. People in rural area are mainly involving in agriculture and related occupations and have more physically active lifestyle and healthier eating pattern.

Problems associated with NCDs can be reduced by early diagnosis and proper management [3]. Therefore, disease-surveillance system remains an integral part of the battle against growing menace of NCDs [11]. Such system will provide the vital information on the magnitude of the problem and it will in turn help to implement necessary preventive strategies [11]. Furthermore, regular surveillance system would provide useful information on disease pattern which can be used to educate the general public and political leadership. Unfortunately, disease-surveillance system remains inadequate particularly in resource-scarce countries like Sri Lanka [12].

Periodic surveillance of NCDs was initiated by Ministry of Health, Sri Lanka with the help of World Health organization (WHO) [13]. However, these surveillances are primarily face-to-face household interviews, and they are both costly and time-consuming [13‐15]. Alternatively, surveillance using telephone interviews has become increasingly used in health surveys, particularly in the developed countries for screening and monitoring of NCDs (e.g. the CDC Behavioural Risk Factor Surveillance System in the US and the Telephone-based Surveillance of Risk and Protective Factors for Chronic Diseases in Brazil [14, 16]. Compared to face-to-face household interviews, telephone surveys are cheap, faster and required less man power to conduct. Furthermore, telephone surveys can be performed more frequently and hence disease trends can be identified early. In addition, telephone surveys give more freedom to access the respondents, regardless of where they live and what they do [15]. Compared to Face-to-face interviews, telephone surveys are less intrusive, warrant greater anonymity and are more convenient for the respondents [15]. It is also an appealing alternative to surveillance involving face-to-face interviews, especially when the access to certain geographical areas is physically challenging.

Even though, telephone surveys tend to be one of the most cost-effective ways of surveillance of NCDs, the main limiting factor is selection bias based on the availability of telephone lines. In Sri Lanka, around 10% of households (2.6 million) have fixed landlines, where as in Galle district, around 12.5% of households own fixed landlines in 2015.

It is a known fact, that people, who own land telephone connections, are comparatively rich and enjoy better quality of life than those without telephone connection. Therefore, it is likely that socially deprived people will be left out from the survey if telephone interviews are used as the only method of surveillance. In addition, there is significant proportion of individuals in the community with undiagnosed NCDs (e.g. diabetes, hypertension) [1, 17]. All these individuals with undiagnosed NCDs will be left out if telephone surveys are used. Apart from the availability of telephone lines, quality of the survey also depends on many other factors such as individuals’ response rate, respondent’s attitudes and their knowledge. Even with these limitations, telephone surveys may still have the potential to be used as a screening method for NCDs particularly in resource poor settings.

There were no previous studies done in Sri Lanka to assess the feasibility and reliability of telephone surveys. Therefore, the present study was designed with main aims of identifying the feasibility and potential challenges of applying telephone surveys for NCDs screening. In addition, the prevalence of self-reported common NCDs (diabetes mellitus, hypertension, ischemic heart disease etc) was assessed based on telephone interview method in this study.

Figure 1 shows household recruitment results and the response rate pattern. Out of 975 of households selected for the study, approximately half (471/975) agreed to participate the study, achieving a response rate of 48.3%. Of the households not included for further analysis, majority did not answer (78%) the phone calls made, despite three contact attempts over three different times. Around 12% were not included as they were not willing to divulge their personal or health related information over the phone. Another 5% disconnected the phone without giving any reasons. Remaining 8% declined to participate the study due to their busy schedules (Fig. 2).

For the final analysis, 1470 adults living in 471 households were included. Baseline and demographic characteristics of the study sample are depicted in Table 1. Respondent ages ranged from 20 to 103 years with a mean age of 50.5 years (SD 17.6 yrs.); 31.2% were aged 20– 39 years, 27.6% were 40-59 years and 34.8% were 70 years or older. Males comprised 46.3% of the study sample with a mean age of 50.2 years (SD 17.2 years). Approximately 20% males and females were “unmarried” and there were higher percentage of female widowers (11) compared to male widowers (1).

The main aim of this study is to assess the potential use of telephone survey in epidemiological health research in Sri Lanka. Despite some concerns over the validity of self-reported data collected via telephone surveys, the use of TS in epidemiological research has grown worldwide [6, 12, 16, 17]. As the telecommunication is growing, with its near-ubiquitous penetration and use in most part of Sri Lanka [22], it is interesting to know whether TS can be used for health related research in Sri Lanka. As far as we are aware, this study is the first study done in Sri Lanka to assess the potential use of TS in an epidemiological health study. Notable findings of this study were, (1) there was an average response rate, with nearly 50% agreeing to participate the study; (2) alcohol consumption was common and predominately seen in males particularly in the middle age group; (3) smoking was exclusively seen in males; (3) young individuals were less likely to engage in health promotion behaviours (regular exercise, salt restriction) than older individuals; (4) self-reported prevalence of diabetes and hypertension was comparable to self-reported prevalence observed in face-to-face studies.

Response rate is one of the most important parameter of efficiency and accuracy of telephone surveys [23]. The overall response rate (48%) observed in our study was comparable to the response rates reported in many other health studies. [12, 14, 15, 24, 25] Even though, response rates for telephone surveys can vary depending on many factors, overall responses rates in many countries have been steadily declining for the last two decades. Long established population health surveys such as U.S. Behavioural Risk Factor Surveillance Study (BRFSS) and the University of Michigan’s Survey of Consumer Attitudes reported declining response rate in recent past. [25, 26]. The trend of declining response rate is likely to be due to a confluence factors, such as busy lifestyle, change of attitude, changing patterns of when household residents are at home, and refusals due to confidentiality issues [24]. In our study, only 22% of non-respondents actively declined to participate the study and the majority (78%) did not answer the phone perhaps the household owners were not at home at the time of calls. Due to practical reasons, we were not able to contact non-responders during after-hours. We believe that we would have achieved much higher response rate if the households were contacted during after-hours.

Alcohol use in our study was 9.1% and it was much lower than the reported alcohol use in the previous Sri lankan studies [13, 27]. A study conducted by Katulanda et al. in 2006 showed higher alcohol usage (23.7%) [27]. In this study, alcohol use in males and females was 48% and 1.2% respectively as compared to 19.2% and 0.4% of males and females in our study. Therefore, our study which was carried out roughly 10 years later showed an overall decrease of alcohol use among both males and females. However, a more recent island wide study revealed higher alcohol consumption, with 39.6% of males and 2.4% of females taking alcohol [13]. The apparent difference in alcohol consumption in our study and the previous studies could have been due to socioeconomic status (SES) of the study populations, the criteria used to define alcohol intake and interview methods. As our study included only those who possessed land telephone lines, our study populations was likely to have better SES than the general population Many studies had shown higher alcohol use by people of low SES than those of higher SES [28, 29], the study subjects in our study was likely to be less socially deprived than those in general population in Sri Lanka.

In contrast to alcohol drinking, tobacco smoking was comparatively low, with only 3.5% of individuals were reported to be active smokers. Similar to alcohol use, tobacco smoking observed in this study was also lower than what was reported in recent past [13, 30]. Survey conducted by Ministry of Health, Sri Lanka in 2015 showed overall smoking prevalence of 10.2%, with significantly higher (19.2%) smoking prevalence in males [10]. In comparison, smoking prevalence among males in our study was 7.5%. Apparent difference in smoking observed in our study and the previous studies could have been due to the comparatively affluent population that we studied, however, the observed difference could be due to the effect of telephone based interview method used in our study. Many studies had shown that telephone surveys generally underestimate cigarette smoking rates and therefore, it is possible that the observed smoking rate of 3.1% in our study was lower than the true prevalence of smoking. [31, 32]

Self-reported prevalence of diabetes in our study was 16.3% with no significant gender difference. This figure is alarmingly high considering the fact that true prevalence of diabetes could be 25 to 50% higher than the self-reported prevalence of diabetes [33]. In an island wide survey conducted in 2014 in Sri Lanka showed self-reported diabetes prevalence to be 7.6% which is almost the half of what we found in our study [13]. Our study, being the most recent and comprising more affluent population by virtue of selection based ownership of telephone lines, may explain the differences between the two studies. It is known fact that diabetes is more prevalent in urban areas in Sri Lanka. In 2008, Katulanda et al. reported a diabetes prevalence of 16.4% in an island-wide study [34] and in 2013,Pinidiyapathirage et al. reported 20% diabetes prevalence in urban setting [35]. Therefore, we assume that the true prevalence of diabetes in our study sample could be as high as 20-22%. As expected, the prevalence of diabetes was higher in older people, with over 28% diabetes prevalence seen in older individuals (over 60 years) as compared to 2.2% in the age group of 20-39 years. The similar rising trend of diabetes with age was reported in many previous studies [36, 37], which perhaps explains the increase in prevalence of diabetes in recent decades.

Hypertension was the most prevalent NCD in our study and was observed in 17.3% of study population with slightly more female preponderance (19.7%). The prevalence of self-reported hypertension in our study was higher than the reported prevalence of hypertension in previous studies [13, 38]. In 2014, an island wide study, involving face to face interview method, showed lower prevalence of self-reported hypertension (6.95%) [13]. However, many other studies had shown significantly higher prevalence of hypertension in Sri Lanka [38, 39]. In 2006, a study conducted in four provinces reported 18.8% and 19.3% of males and females respectively having hypertension [38]. In another community based study involving over 2900 individuals in an urban area showed comparatively higher prevalence of hypertension (30.4%) [40]. In this study, around 30% of hypertensive individuals had previously undetected hypertension [40]. Therefore, it is likely that true prevalence hypertension in our study may be 30% higher than the observed prevalence.

Roughly around 15% of individuals in the study sample either at present or in past were on statin, to be classified as having dyslipidaemia. Unlike diabetes mellitus and hypertension, that were more prevalent in age category of >80 years, dyslipidaemia was more prevalent in age category of 60-79 years. Even though, there were no previous Sri Lankan studies to compare, a large Indian study conducted in a similar setting revealed a similar prevalence of hypercholesterolemia [41]. Furthermore, most patients (>90%) with dyslipidaemia in our study had one or more other NCDs such as diabetes and hypertension and only 8.4% had isolated dyslipidaemia.

This study was conducted primarily to assess the feasibility and potential challenges of using telephone surveys for health related epidemiological study in Sri Lanka. The main strength of our study is the use of a novel, inexpensive, efficient, health-survey methodology that has the potential to improve epidemiological research in Sri Lanka. In addition, inclusion of a random, representative, comparatively large number of adults in the study is also a strength. Despite these strengths, this study has several limitations and therefore, interpretation of the finding of this study should be done in the context of these limitations. Despite we made strenuous efforts to directly interview all adult members in the households, data related to some individuals (around 10%) were obtained from next of kin. Therefore, some of the information (such as tobacco smoking and alcohol use) obtained from next of kin may not be 100% accurate, and hence, can potentially underestimate the true prevalence of these factors. Another important, but unavoidable limitation is the use of self-reported telephone-based information to assess the prevalence of all NCDs. Many studies have shown that the true prevalence of NCDs such diabetes mellitus, dyslipidaemia and hypertension can be significantly higher than the self-reported prevalence of NCDs [42, 43]. Another potential limitation of this study is the sampling bias. By virtue of selecting individuals based on the ownership of land telephones can lead to selecting more affluent individuals to the study and excluding socially deprived individuals. Even with these limitations, our findings support the potential use of telephone surveys in health related research and use of self-reported information for assessment NCDs in Sri Lanka. Since the NCDs are the leading cause of morbidity and mortality in Sri Lanka at present, the pattern and trends generated, along with other indicators, will be useful to evaluate NCDs and other health related parameters in Sri Lanka.

In conclusion, this study has shown an acceptable and average response rate for telephone based interview in Sri Lankan setting. Overall, prevalence of main NCDs obtained in this telephone based survey showed comparable prevalence, and trends to studies used face-to-face interview method and hence can support the use of telephone-based information to monitor the prevalence and the trends of main NCDs in Sri Lanka.