Quality outcome of diabetes care during COVID-19 pandemic: a primary care cohort study

Aim Management of diabetes care can be affected by COVID-19 pandemic control measures. This study aimed to determine the impact of the pandemic, during 17.03.2020–16.03.2021, on quality outcomes of diabetes care in general practice in Switzerland. Methods In this retrospective cohort study, diabetes mellitus patients (≥ 18 years) with at least one consultation at a general practitioner, during 17.03.2018–16.03.2019 (cohort 1) and 17.03.2019–16.03.2020 (cohort 2) were included and followed-up for two years. Quality indicators and outcomes of diabetes care, at patient and practitioner level, were compared before and during the pandemic. Logistic regression was performed to identify patient’s risk factors for dropout from follow-up. Results Data from 191 practices, 23,903 patients, cohort 1 and 25,092 patients, cohort 2, were analyzed. The fraction of patients lost to follow-up, attributable to the pandemic, was 28% (95% confidence interval: 25%, 30%). During the pandemic, compared to the previous year, regular measurement of weight, HbA1c, blood pressure and serum creatinine were less frequent and less patients per practitioner reached HbA1c and blood pressure target outcomes. Factors associated with continuity of care during the pandemic were: patient age 41–80 years, longer diabetes duration, diagnosis of hypertension or dyslipidemia, influenza vaccination during the last year. Risk factors for dropout were age > 80 and receiving only insulin as anti-diabetic medication. Conclusion A considerable quality reduction in diabetes mellitus care could be observed during the pandemic. Though the most vulnerable patients were not the most affected by the pandemic, key factors that might reduce dropout from follow-up were identified. Supplementary Information The online version contains supplementary material available at 10.1007/s00592-022-01920-0.


Introduction
During SARS-CoV-2 pandemic, management of chronic non-communicable diseases (NCDs) such as diabetes, hypertension, dyslipidemia can be affected in several ways.
Managed by Massimo Porta.
Even in the absence of an overload of COVID-19 cases, disease control measures, such as lockdown, quarantine, restrictions of public and private transport or fear of infection might have an impact on accessibility of health care [1]. In Switzerland, non-urgent patient care was suspended during lockdown (17.03.2020-26.04.2020) [2].
Impact of COVID-19 pandemic on quality of diabetes care, as promoted through the Quality and Outcomes Framework (QOF) [17,18], has been scarcely measured [19]. Processes and outcome indicators for type 2 diabetes patients, according to the Italian guidelines, were also compared between 2019 and 2020 [20].
This study aimed, first, to assess the impact of SARS-CoV-2 pandemic, during 17.03.2020-16.03.2021, on quality indicators and outcomes of diabetes care, based on [17,18] and adapted for primary care in Switzerland [21][22][23]; second, to identify factors associated with patient dropout from follow-up during the pandemic.

Study design and population
This retrospective cohort study used the database from FIRE project (Family Medicine ICPC (International Classification of Primary Care) Research Using Electronic Medical Records) [24,25] which acquires medical data from general practitioners in Switzerland.
Adult patients (≥ 18 years) with diabetes mellitus and at least one encounter at a general practitioner during baseline year: 17.03.2018-16.03.2019, cohort 1, and 17.03.2019-16.03.2020, cohort 2 (pandemic-exposed), were included and followed-up respectively before the Swiss lockdown, until 16.03.2020, and from Swiss lockdown until one year, 16.03.2021.
Diabetes mellitus diagnosis was based on one of the following conditions from whole patient history: i) at least two measurements HbA1c ≥ 6.5% (48 mmol/mol), as recommended [26]; ii) prescription of any anti-diabetic medication (Anatomical Therapeutic Chemical Classification System (ATC) [27] code A10); iii) International Classification of Primary Care 2nd edition (ICPC-2) [28] diagnosis Code T89 or T90. As Glucagon-Like-Peptide (GLP)-1 receptor agonists and Sodium-dependent Glucose Transporter 2 (SGLT-2) inhibitors could be prescribed for other reasons than diabetes (obesity and congestive heart failure or chronic kidney disease, respectively) patients treated exclusively with them were included if in addition at least once HbA1c ≥ 6.5% (48 mmol/mol).
Postal code of the physician practice was used to identify urban, suburban, rural areas [29].
Time from first diabetes diagnosis to study start was categorized as: first diagnosis during baseline, first diagnosis < 1 year, 1-5 years and > 5 years.
Insulin-dependent, non-insulin dependent and unknown was a proxy for diabetes type (Online Resource 1 Table 2).
Other medications relevant to diabetes care were: Aspirin, statins, RAAS-inhibitors. Data on influenza vaccination was also reported (Online Resource 1 Table 3).
New or not expired prescriptions, during the observation period, were included. For prescriptions without defined stop dates, a validity of 365 days was supposed, as most of diabetes patients with a particular treatment, had the same prescription in the following 12 months [31].

Statistical analysis
Baseline characteristics of each cohort: during 17.03.2018-16.03.2019 and 17.03.2019-16.03.2020 respectively, were described as number and percentage, N (%), for categorical or binary variables and as mean standard deviation (SD) for continuous variables. χ 2 -test, for categorical or binary variables, or t-test, for continuous variables, were performed for cohort comparisons or baseline and follow-up comparisons within cohorts.
Trends of all HbA1c values, weekly averaged over patients by cohort, were shown graphically.
For each cohort, from baseline to follow-up, differences between average values, for laboratory measurements, or between proportions, for quality indicators, were reported with 95% confidence interval (CI).
A subgroup analysis of outcome indicators during each year, for patients included in both cohorts, was also performed.
Quality indicator results were shown through a dumbbell plot or connected dot plot. At practice level, the median patient proportion, in each cohort, for each indicator, during baseline and follow-up periods, was reported with the interquartile range [IQR] and represented through error bar plots.
Population attributable fraction (PAF) [32] was used to compare proportions of cases in the two cohorts, with complementary outcome (dropout from follow-up, not reaching quality target …) during follow-up, considering cohort 2 being pandemic-exposed. PAF was shown graphically through a bar chart with 95% (CI) error bars.
To identify risk factors for dropout from follow-up during the first year of pandemic, for cohort 2, unadjusted and multivariable-adjusted mixed logistic regression models were performed. Random effects were considered at practice level, to correct for correlation between patients followed by the same practice. Predictors in multivariable analysis were selected with a stepwise backward approach, starting from a full model including all variables, not correlated among them, with p < 0.2 in univariable analysis. Results of regression analysis were reported as odds ratio (OR) (95%(CI)). Multivariable analysis results were represented through an odds ratio plot.
Since microalbuminuria was scarcely reported PAF was not evaluated.
LDL quality indicators improved, from baseline to follow-up in each cohort, though far from the ideal threshold (dashed line). As most patients did not reach outcome targets, PAF for these outcomes were not evaluated.

Factors associated with dropout from follow-up during the pandemic
Online Resource 1 Table 8 and Fig. 5 reported results of univariable and multivariable analysis of factors associated with dropout from follow-up during the pandemic. From multivariable analysis, protective factors against dropout were:  Dashed line represented the quality reference area, or threshold, for each indicator. Attributable fraction in the exposed (cohort 2) and in the population, PAF (two cohorts), the latter with 95% confidence interval (CI), were reported in the barplot (right side). They were calculated, for the complementary outcome of each indicator in the follow-up year, as difference between the respective proportion and the proportion of unexposed cases in cohort 1. In legend, pandemic meant attributable to the pandemic exposure. Abbreviations: n: number of reported measurements per patient; HbA1c: Hemoglobin A1c; BP: blood pressure; LDL: low density lipoprotein

Summary
In this study, the impact of COVID-19 pandemic on quality and outcomes of diabetes care was evaluated. The main findings are: (i) 28% of total dropout from follow-up during the observation period was attributable to the pandemic; (ii) the proportion of patients with regular measurement of weight, HbA1c, blood pressure and serum creatinine decreased during the pandemic compared to the previous year; (iii) at practice level, the proportion of patients reaching HbA1c and blood pressure target decreased during the pandemic compared to the previous year; (iv) factors associated with continued care during the pandemic were: patient age 41-80; longer diabetes onset; diagnosis of hypertension or dyslipidemia; influenza vaccination in the previous year. Risk factors for dropout from follow-up were age > 80 and receiving only insulin as anti-diabetic medication.

Strengths and limitations
A strength of this study is the large database used whose validity is also supported, being the distribution of age, gender, as well as prescription proportions of anti-diabetic medications in agreement with Swiss health care settings [4,34].
Conversely, there are some limitations. First, data quality had impact on baseline characteristics differences between the two cohorts. Moreover, we could not reliably distinguish between types 1 and type 2 patients or even be sure to have included all diabetes patients. Second, the database used (FIRE) only includes general practitioners but some of them could have a double specialty in endocrinology and general internal medicine. Moreover, some patients might have been followed by general practitioners and endocrinologists at the same time, having made their laboratory analyses at the general practitioner, but received their prescription at the endocrinologist, or vice versa. Third, we had no information about patient adherence or tolerance to treatment that could have influenced prescribing decisions. Fourth, data concerning patient-physician contact (face-to-face consultation, telephone or video-call) was unrecorded. Fifth, reason for dropout from follow-up (death, hospitalization/ institutionalization, change of physician) was unknown. Sixth, prescriptions without defined stop dates might have been overestimated, supposing a validity of 365 days. Seventh, we could not examine the impact of socio-economic or life-style variables. Last, information about COVID testing or infection was missing. This could have affected dropout, increased comorbidities during the pandemic, glycaemic control or prescriptions.

Pandemic and dropout from follow-up
During 17.03.2020-16.03.2021 the dropout rate was 14.3% of which 43% attributable to the pandemic. At population level, the pandemic-attributable fraction was 28%. Several studies found a negative impact of lockdown on consultations for diabetes patients: weekly consultations were 17.5% lower than expected without lockdown [35]; during lockdown 49% of patients did not consult their general practitioners [3], similar to [36,37], while in India the majority reported no access to healthcare services [38]. In a large cohort of around 250′000 type 2 diabetes patients in Italy, an overall reduction of 24% in follow-up visits was observed during 2020, compared to 2019 [20]. As no study evaluated the proportion of dropout cases attributable to the pandemic, in a year time after lockdown started, our findings are not directly comparable with the existing literature.

Pandemic and anti-diabetic prescriptions
During the pandemic and compared to the previous year, the proportion of patients with medications increased in our study and for SGLT-2 and GLP-1 a significant effect was attributable to the pandemic. Other studies observed: an increase in insulin [39] and both insulin and OAD medications [40,41], during the first month of pandemic as compared to the year before; a decrease in OAD during the first four months of the pandemic [42].

Pandemic and quality indicators of diabetes care
According to our findings, the reduction in measurement counts, for all primary care patients, was more pronounced than the reduction in consultation counts [35], though our results were at patient level and not at consultation or measurement counts. Marked reductions in the rate of health checks of type 2 diabetes patients, between March and December 2020, were highlighted [19,20]. Accordingly, quality at patient level declined, during the pandemic, in particular for the number of patients with HbA1c recorded (17.6% absolute difference) and for the number of patients with blood pressure recorded (16.6% absolute different). However, differently from the literature, we were the first reporting the pandemic-attributable fraction for not having HbA1c and blood pressure records: PAF 19% and 14%, lesser than the one for dropout from follow-up, 28%.
We considered a larger time frame compared to these studies. Being HbA1c the most important variable of diabetes care, we analyzed weekly averaged measurements during baseline and follow-up, by cohort, evidencing around 0.1% (1.1 mmol/mol) higher values, one month after the lockdown. That means higher glucose levels during lockdown, as HbA1c correlates with mean glucose level in the previous 8-12 weeks [43], though with minor clinical impact, as after five months HbA1c returned to the previous year's level.

Factors associated with dropout from follow-up during pandemic
To our knowledge, this is the first study to assess association of diabetes patient characteristics and medications with dropout from follow-up during a year from the lockdown. Reduction in follow-up visits in type 2 diabetes patients was independent of age, sex, and educational level [20]. Other studies investigated the reasons to avoid or postponed the visit during lockdown [3,40]. Differently from [20], we found that dropout of youngest patients, age ≤ 40 years, was the most affected by the pandemic, PAF 37%(30, 43)% though the oldest ones, age > 80 years, had the highest risk of dropout, after correcting for confounders. However, this higher risk was not attributable to the pandemic, since dropout of oldest patients was the less affected, PAF 19%(13, 24)%, in line with [35,44]. Therefore, since comorbidities were associated with regular care, the most vulnerable patients remained the main focus of primary care despite the pandemic.

Implications for research and/or practice
This study showed a decline in diabetes mellitus quality care during COVID-19 pandemic between 17.03.2020-16.03.2021, especially when facing HbA1c ≤ 7.0% (53 mmol/mol) and blood pressure < 140/90 mmHg. For most indicators, around one third of patients, included in both cohorts, reached the quality outcome during the pre-pandemic years but no during the pandemic. However, there was also a relevant proportion of patients, around one fifth for most indicators, not reaching outcome indicator in every year of observation, suggesting room for improvement in quality of diabetes care, independently of the pandemic. Though the most vulnerable patients (old, with more comorbidities) were not the most affected patients by the pandemic, our finding suggests key factors that might reduce dropout from follow-up of patients with diabetes mellitus. Primary care should have a primary role in guaranteeing continuity of care of these patients in order to prevent long-term adverse effects of the pandemic on diabetes complications.
Acknowledgements The authors would like to thank the FIRE team for their expertise in data quality and support throughout clinical definitions, data extraction and data management. In particular, we would like to thank the Head of FIRE Jakob Martin Burgstaller MD DMD PhD, the Head of FIRE Research Stefan Markun MD and Levy Jäger MD, Giuseppe Pichierri PhD, Fabio Valeri Msc for technical support.
Author contributions All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Stefania Di Gangi. Funding acquisition was managed by Roland Fischer. Supervision was carried out by Andreas Zeller and Roland Fischer. The first draft of the manuscript was written by Stefania Di Gangi and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.
Funding Open access funding provided by University of Zurich. The study received a research grant by KHM (Kollegium für Hausarztmedizin), https:// khm-cmpr. ch/ forsc hungs fonds-gewin ner-2021/. The founder had no involvement in the study design; collection, analysis and data interpretation; writing of the report; or the decision to submit for publication.

Data availability
The datasets analyzed during the current study is available from the corresponding author on reasonable request.

Conflict of interest
The authors declare no competing interests.
Ethical approval As the project did not fall under the scope of the Federal Act on Research involving Human Beings (Human Research Act) (BASEC-Nr: Req-2017-0079), no ethical approval was needed.
Informed consent As established by the Swiss Federal Law (HFG) and by the cantonal ethics committee, informed consent was not applicable because the study used anonymized data irreversibly altered, according also to the European General Data Protection Regulation.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.