The effect of COVID-19 pandemic on diabetes care indices in Southern Iran: an interrupted time series analysis

Diabetes is one of the most prevalent chronic disorders that has reached the epidemic proportions worldwide, exerting a substantial burden on the health care service [1]. Based on the estimates from 2019, 1.1 million children and 463 million adults (aged 20 to 79) worldwide have diabetes, and it is expected to increase to 700 million by 2045 [2]. Furthermore, 12% of Iranian individuals have diabetes, placing a significant strain on country's healthcare system [3].

Like many chronic diseases, the patients with type 2 diabetes mellitus (T2DM) need a long-standing healthcare coverage to prevent micro- and macro-vascular complications [4]. The coronavirus disease 2019 (COVID-19), the most recent infectious disease to go global, was shown to have a common underlying illness across COVID-19 patients that has a discernible effect on the disease's course and prognosis [5‐8]. Besides, a new pandemic – i.e., COVID-19 – affects healthcare system framework in response to the influx of the patients with COVID-19, including the divergence of resources, and bias toward the disease itself, the redistribution of health services infrastructures (health workers, equipment, and facilities), financial barriers, physical restrictions, and public fear of infection. By these inescapable challenges, governments, especially low-to-middle income countries, have faced declines in the health service coverage. Consequently, by emerging these direct and indirect effects, diabetes healthcare coverage might drop dramatically [9], which might increase the burden of diabetes. Hence, it is essential to study the changes in diabetes healthcare indices to find solutions to improve the diabetes care.

People who suffer from the chronic conditions are more prone to infections. Such factors decrease the likelihood of the recovery for COVID-19 patients [10, 11]. The most difficult aspect of a pandemic is continuing to provide normal treatment for chronic conditions [12]. The difficulties of the routine healthcare provision for T2DM patients during COVID-19 pandemic in Iran could be summarized in nine themes as follows: the lockdown of standard outpatient clinics, decreased inpatient capacity, staff shortage, medical facilities shortage, unaffordable medicine, delayed care seeking, limited self-care practice, transport difficulties, and undiagnosed cases/events [13]. During the study period, there was a lockdown for COVID-19 in Iran. We experienced 3 peaks of COVID-19 pandemic. Moreover, the significant changes were observed in the diabetes self-management behaviors, and using the telemedicine and virtual and online visits increased [14, 15].

Some evidence indicates serious damages to cover the chronic diseases’, especially diabetes, preventive and therapeutic services during current pandemic [16‐19]. While in another study, there was no difference between the control and quarantine groups due to body mass index (BMI), plasma glucose, or hemoglobin A1c (HbA1c) levels [20]. However, the supporting data is weak, and no comparable research using Iranian data was performed. We sought to assess the effect of COVID-19 pandemic on diabetes healthcare indices in the health centers under the auspices of Shiraz University of Medical Sciences, Shiraz, southern Iran.

In this ecologic study, we assessed the effect of COVID-19 announcement as a natural intervention to T2DM patients’ healthcare indices. From January 01, 2020 to January 01, 2021, 160,000 T2DM patients who were under cover of Shiraz University of Medical Sciences, Shiraz, Iran in health centers of 29 counties in Fars province, included into the study. The population of Fars province was stable before and after the epidemic.

The target statistics were the totals of all T2DM patients living in the Fars Province, Southern Iran, areas covered by the Shiraz University of Medical Science (SUMS), from 2019 to 2020.As a nature of administrative data, in each visit, diabetes care parameters were electronically recorded in the Integrated Health System (in Persian: “Samaneh Yekparche-ye Behdashti”) by health staff mainly working in primary and secondary healthcare centers (including 29 cities, urban and rural centers). For research purpose, the seasonal aggregated counts for various indices were integrated in the SUMS Department of Health. Individual data were not collected.

The final dataset contained several aggregated variables, including numbers of the patients with T2DM and newly diagnosed patients with T2DM, sex, season, diabetes care-related measurements, such as health worker visits, physician visits, BMI (body mass index) [3 levels, desired level < 25 kg/m2], FBS (fasting blood sugar) [3 levels, desired level of 70–130 mg/dL], and HbA1c (hemoglobin A1c) [5 levels, desired level < 7%], health status and outcome variables, such as hypertension (BP < 140/90 mmHg), uncontrolled hypertension (BP > 140/90 mmHg), microvascular and macrovascular complications, emergency refers to hospital (e.g., sudden complication, dysglycemia, etc.), and the reasons (death, migration, etc.) do not go to the health centers.

Based on the national guideline [21], each T2DM patient should be regularly seen by a health care worker every month and a physician every three months. The health professionals should assess the patient's BMI and mean blood pressure at each visit and record the data every three months. Besides, family doctors should send patients to laboratory facilities to assess HbA1c every 6 and 3 months, respectively, for those with and without glycemic management. The primary result of our study was to assess the immediate secular trend, induced by COVID-19 pandemic announcement, on T2DM healthcare indices. The secondary outcome was to track the seasonal trend of these indices one year before, and after COVID-19 pandemic announcement, using Interrupted Time Series analysis.

The point intervention day was set as February 20, 2020 when the first case of COVID-19 officially announced in Shiraz, seat of Fars province. Statistical analysis was done using the STATA (Stata Corp (2017) Stata Statistical Software: Release 15. College Station, TX: Stata Corp LLC.). We used seasonal values as the unit of time period for interrupted time series analysis as well as aggregated data; hence we could not assess the heterogeneity. Besides, we assumed that the revelation of the COVID-19 pandemic was the only unexpected intervention or event that happened throughout the research period. To test this hypothesis, we performed a non-parametric correlation analysis (Spearman's rho) among three pandemic severity indicators (seasonal counts of the COVID-19 incidence rate, hospitalization rate, and fatality rate) and their corresponding seasonal counts of T2DM care indices. In general, we did not observe any correlation; that is, the T2DM healthcare services changes or recovery during the first year of COVID-19 pandemic did not potentially affected by the pandemic severity (table S18). For data handling, Microsoft Excel data analysis software was used. Data were described as mean (95% confidence interval (CI)). The segmented Poisson regression was selected since each dependent variable was constituted from the counts of care delivered in a season. The interrupted time series models require at least three independent variables [22]:

Beta (β) coefficients indicates trend or level of change in each variable, including β1 for the trend before COVID-19 pandemic, β2 for the early change in level induced by the COVID-19 natural intervention, and β3 for the trend of recovery (or deterioration) during COVID-19 pandemic. For each β coefficients, a P-value ≤ 0.05 indicates that a trend or change is statistically significant.

Approximately a mean 160,000 T2DM patients (61.9% female and 79.5% urban resident) were seasonally observed from 2019 to 2020. A mean of 6,254 individuals (55.7% female, 80.4% urban resident) were newly diagnosed with T2DM (Table 1). The mean prevalence of T2DM was 7.2%; and 6.9 and 7.3% in 2019 and 2020, respectively. Moreover, the mean incidence of T2DM was 30 per 10,000 people with 40.3 and 19.9 per 10,000 people in 2019 and 2020, respectively.

The mean seasonal number of all diabetes care indices was decreased in 2020 comparing 2019, which are 49.9% vs. 35.8% in health worker visits, 43.9% and 28.9% in physician visits, 49.5% vs. 35.3% in BMI assessment, 38.7% vs. 25.7% in FBS assessment, and 23.9% vs. 14.1% in HbA1c assessment. This consistent reduction in the aforementioned indices was found for every pair of comparable seasons in 2019–2020 (Table 2). Besides, seasonal crude figures for the health condition of T2DM patients in 2020–2021 are included in Table 2.

Table 3 shows trend, that induced by the COVID-19 pandemic, of diabetes care and outcomes. Both seasonal incidence (β1 = -0.028, P = 0.689; β2 = -0.685, P = 0.012; β3 = 0.062, P = 0.610) and prevalence (β1 = 0.030, P = 0.859; β2 = -0.052, P = 0.924; β3 = -0.012, P = 958) of T2DM was almost stable (not statistically changed) during the study period (Fig. 1 A, B).

A significant decrease in the number of visits by health workers and physicians was observed before COVID-19 pandemic (β1 = -0.046, P < 0.001; β1 = -0.034, P < 0.001) which got worse by the first month of COVID-19 pandemic (β2 = -0.245, P < 0.001; β2 = -0.292, P < 0.001); but, it was recovered by following months (β3 = 0.112, P < 0.001; β3 = 0.053, P < 0.001) (Fig. 1C, D). The exact pattern was observed for number of BMI, FBS and HbA1c assessments (Table 3, Fig. 1 E- G).

A stable trend was observed for the proportion of T2DM patients with HbA1C < 7% (β1 = 0.023, P = 0.751; β2 = -0.125, P = 0.614; β3 = 0.010, P = 0.919) and controlled HTN during study period (β1 = -0.099, P = 0.414; β2 = -0.312, P = 0.492; β3 = 0.202, P = 0.294). But, a significant increasing pattern was observed in number of newly diagnosed diabetes complication before COVID-19 pandemic (β1 = 0.087, P < 0.001), but decreased after announcing the first cases of COVID-19 (β2 = -0.566, P < 0.001), and also by the next months (β3 = -0.147, P < 0.001) (Table 3).

Trend in number of refer to hospital emergency wards was stable and unchanged before the COVID-19 pandemic (β1 = 0.005, P = 0.608). By the emergence of COVID-19, this figure dropped significantly (β2 = -0.401, P < 0.001). Then, it increased in the next months (β3 = 0.106, P < 0.001) (Table 3). For interest of the readers, results of further analyses on the changes in diabetes care indices are yielded in two separate Supplementary files (file 1 contains 18 tables and file 2 contains 33 figures). Supplementary table 1 (table S1-S3) are description of population, epidemiological and diabetes care indices stratified by gender and season. The optimum and unsatisfactory health care indicators, divided by gender and season, are described in Table S4-6. Referring (refer to level 2 and hospital emergency ward) and not going to the health centers indices are described in Table S7-9 and are broken down by gender and season. Table S10-17 are Interrupted Time Series (ITS) analyses of just-mentioned indices based on the gender and season. Table S18 is the correlation of seasonal T2DM care numbers and seasonal rates of COVID-19 incidence rate, hospitalization and mortality rate during the first year of the COVID-19 pandemic in Fars province, Southern Iran. Supplementary Fig. 1 (figure S1-S33) are plots of ITS analyses on diabetes indices according to gender and season.

The healthcare service coverage is supposed to be affected by the outbreaks and epidemics, generally experiencing an early secular change (“Shock effect”) followed by a recovery trend toward pre-intervention levels (“System reaction”). The resources, social environment, and population behaviors all have a significant role in determining this recovery period's features, such as the recovery gap. According to reports, there is a declining trend in the coverage of diabetic healthcare services throughout the present COVID-19 pandemic [16‐19]. Similarly, a significant decrease in the diabetes healthcare indices was found for number of visits by health workers and physicians, the number of BMI, FBS and HbA1c assessments, and number of refer to hospital emergency wards with the shock on the announcement of COVID-19 pandemic. Nonetheless, almost all of the measured indices started to significantly recover after a gap of one to two months. These changes in diabetes care visits and assessments might be associated with redistribution of health staff and facilities, lockdown policy, community fear and phobia (believes), and performance of mass media. The health surveillance system advised T2DM patients to self-monitor their blood sugar, live a healthy lifestyle, be instructed to drink enough water, exercise during lockdowns, use alternative strategies like telehealth, virtual and digital services, and have regular and continuous follow-up with efficient educational programs in order to improve the coverage of diabetes care [13].

Though only 4 times before the interruption is insufficient for trend analysis, a significant decreasing trend was present in number of visits by health workers and physicians, as well as BMI, FBS and HbA1C measurements before COVID-19 pandemic. One can emphasize on some weakness in our primary and secondary health system probably related to the weakness in the health economy, the disproportionate distribution of medical services and professionals, and health system bias toward tertiary care and. Regarding this situation, the COVID-19 pandemic's converging force might exacerbate this healthcare delivery deficiency. The patients with non-communicable illnesses, particularly those with T2DM, were more at risk when the majority of healthcare resources were diverted to the treatment of COVID-19 patients. Besides, many patients with such chronic conditions avoided their direct exposure to medical centers because of the risk of infection or the widespread fear and rumors. Furthermore, restrictions and quarantine measures established by the governments could disrupt self-care activities, insufficient clinical support and less monitoring of these patients [13, 23].

In theory, these changes would affect glycemic control and diabetes outlooks. In two reports from India and Singapore, no significant change in HbA1c and BMI mean levels before and after COVID-19 pandemic [24, 25]. These findings might be explained by the healthy lifestyles (healthy eating and exercise) and a brief observation period [26, 27]. In our study, the mean seasonal number of T2DM patients with controlled BMI, FBS, HbA1c, and blood pressure had decreased in 2020 compared to 2019 for each pair of similar seasons. However, these decreases may simply reflect a lower number of records in 2020 rather than an actual decrease in the number of patients with glycemic control following the COVID-19 pandemic. Hence, we showed that the proportions of HbA1C < 7% and controlled HTN did not significantly changed with the onset of the COVID-19 pandemic or in the following seven months. Noticeably, a portion of these secular trends should be attributed to the diabetes surveillance system purposely decisions. For example, during the pandemic, telecommunication and self-monitoring were suggested. We showed that the number of T2DM patients without any referral (excluding died patients) was statistically increased by starting COVID-19 pandemic; but the number of mortality-related do not go to the health centers and newly diagnosed diabetes complication were not statistically increased. Hence, it would be plausible to observe a secular trend in healthcare coverage without any significant change in outcomes. Although this statement should be considered with caution. For instance, it is well recognized [9, 28, 29] that mortality and mortality-related avoidance of health facilities is dramatically increasing. Or, Khader et al. [19] showed that the majority of T2DM patients who underwent home monitoring during COVID-19 epidemic had uncontrolled blood glucose levels. Besides, the mean incidence of diabetes was 19.91 per 10,000 people in 2020, which was lower than the mean of 40.27 per 10,000 people in 2019 (40.27 per 10 thousand). It might be related to the less disease screening and detection during COVID-19 pandemic, as well as the introduction of unusual nationwide policies (i.e., social distancing measures- Applying the limitations, implementing lockdowns, and remaining at home may interfere with the procedure for locating new patients. While the average prevalence of T2DM in 2019–2020 was 7.2%, our research team recently released a study in the province of Fars where we evaluated the prevalence of T2DM to be 7.2% and 8.74% in 2015 and 2016 [30].

The major strength is its’ large population (~ 160,000), while most of in-line studies are conducted on a substantially lower number of subjects. The major limitation was the lack of minimum time unit required for Interrupted Time Series analysis. Penfold and Zhang have stated that a minimum of eight events is required for before, and after the intervention to achieve robust result. Our data were inherently seasonal, which yielded four time-units for both before and after the intervention. Although interrupted time series analysis can be done with a smaller number of time units [31]. Furthermore, our data were inherently administrative – not for research purpose – that might be inevitably defective during the record process by health staff.

We performed a large-scale ecologic study in southern Iran to assess the change in diabetes healthcare indices induced by the COVID-19 pandemic. In conclusion, we discovered that the mean incidence of diabetes decreased from 2019 to 2020. T2DM prevalence was 7.2% on average in 2019–2020.Besides, the healthcare coverage sharply declined with the announcement of COVID-19 pandemic, due to the shock effect. However, soon started to recover despite a gap of one to two months. Our data showed no significant changes in the diabetes outcomes, which might imply on the health surveillance system reaction. Finally, it is advised to adopt alternate tactics like telemedicine, digital services for consultations, self-management, remote monitoring, etc. to continue diabetes treatment while facing a pandemic status in the future. Besides, with a consistency in the health surveillance system, it would be more plausible to tackle issues confronting with unexpected massive challenges. Maintaining healthcare access for all should be a key priority in public health.