Advertisement

Quality of Hypertensive Patients’ Electronic Health Records in Specialized Cardiological Centre: 6-Year Trends

  • Anna Semakova
  • Nadezhda Zvartau
  • Ekaterina Bolgova
  • Aleksandra Konradi
Conference paper
Part of the Communications in Computer and Information Science book series (CCIS, volume 674)

Abstract

Electronic health records (EHRs) have the potential to form the basis for a personalized approach to patient management, deliver high-quality care, and make the healthcare system more efficient and safer. Finding and studying the possible trends and long term changes of individual results in stored medical data may facilitate selection of an optimal treatment plan. Moreover, guidelines on disease management usually include only shifted results of clinical trials that are poorly generalized to routine clinical practice. In numerous EHR-related errors have been described, such as unstructured data, missing data, and incorrectly entered data. This study aims to assess the quality of hypertensive patients’ EHRs during 6 years after the implementation of EHRs in the specialized cardiological centre. The quality of patients’ EHRs was estimated by the completeness and consistency of stored data. We compared information entered into EHRs with diagnostic algorithms recommended by hypertension management guidelines. The results demonstrated the incompleteness and inconsistency of information in EHRs on risk factors, diabetes mellitus (DM), and subclinical organ damage. An assessment of six-year trends showed that the quality of data decreased in parallel with increase of workload of the clinic (estimated by the number of primary visits). Results indicate the urgent need for an action plan to resolve the problem of data incompleteness and inconsistency. Integration of specially designed clinical decision support system (CDSS) considered as a possible decision promoting an increase of EHRs quality. This study is part of a larger project aimed to develop of complex CDSS on cardiovascular disorders for medical research centre.

Keywords

Electronic health records Risk factors Diabetes mellitus Subclinical organ damage Completeness and consistency of information Learning curves Statistical data analysis 

1 Introduction

Electronic health records (EHRs) may promote the introduction of a personalized approach to patient management and make the healthcare system more efficient. Finding and studying the possible trends and long term changes of stored patient medical information facilitates the selection of an individualized optimal diagnostic treatment plan. EHR in its ideal form for patient care is a longitudinal record of patient health information generated by multiple encounters in any care delivery setting [2, 9]. If an EHR is fully implemented, reuse of EHR data may be extremely helpful in supporting clinical research [5, 12, 15]. Integration of clinical research data with patient clinical data may provide a better understanding of true individual health status, clinical trials feasibility, etc. [6].

However, there are many obstacles to be overcome in using EHRs for clinical research [7, 14, 16]. Data quality (completeness, consistency) is another challenge for the reuse of EHR data [3, 4, 8]. A number of EHR-related errors, such as unstructured data, missing data, and incorrectly entered data, lead information incompleteness and inconsistency. These limitations currently prevent the optimal use of EHR patient data and information, and impede the advancement of medical research, the improvement of healthcare, and the enhancement of patient safety [1, 17].

In our research, we used a methodology for estimating the quality of patient EHRs, based on an assessment of the completeness and consistency of stored data. We assessed the completeness and consistency of data according to predefined in hypertension guideline risk factors and subclinical organ damage [13]. These data should be present in all hypertension patients as they are necessary for evaluation of prognosis and choice of optimal treatment strategy.

2 Current Study

2.1 Data Description

In our study, we used the depersonalized EHRs of patients referred to a specialized cardiological centre due to uncontrolled arterial hypertension (AH) during 6 years period (from 2010 to 2015). EHR represents the case history of a patient and contain information about all patient visits to the cardiological centre, complaints, results of examination, and prescribed investigations and treatment. We included only patients with an initial diagnosis of AH, which means that high blood pressure was the main reason for referral to the cardiological centre and a hypertension management algorithm had to be applied to develop a diagnostic and treatment plan.

In the current study, we focused on information filled by the treating physician during primary outpatient visit. The information in EHRs for outpatient visits contains following data:
  • general information about patient such as age, gender, etc.;

  • anamnesis vitae and anamnesis morbid, medical examination by a doctor: anamnesis vitae with family history of premature cardiovascular diseases (CVD), information about bad habits (smoking and drinking), complete history of the development of current condition, blood pressure level, height, weight, results of medical examination by a doctor, etc.;

  • recommendations: diagnosis and prescriptions;

  • various medical events or procedures: test results, instrumental investigations and other procedures.

2.2 Current Hypertension Guideline Criteria

According to current guidelines, 10-year risk of cardiovascular death should be estimated in each hypertensive patient and reassessed regularly. The risk is predicted by the severity of hypertension (the increase of blood pressure level), presence of risk factors, subclinical organ damage or clinically manifest cardiovascular and other diseases (DM, end-stage renal disease). Risk factors include blood pressure level (>140/90 mm Hg or intake of antihypertensive medications), male sex, age (men ≥ 55 years; women ≥ 65 years), dyslipidemia (lipids disorder), high glucose level, obesity (body mass index (BMI) ≥ 30 kg/m2), smoking and family history of premature CVD (men aged < 55 years; women aged < 65 years); subclinical organ damage – left ventricular hypertrophy (LVH) by ECG and ECHO data, estimated Glomerular filtration rate (eGFR), microalbuminuria (30–300 mg/24 h) or albumin–creatinine ratio (30–300 mg/g; 3.4–34 mg/mmol), DM – glucose and glycated haemoglobin (HbA1c) [11].

2.3 Quality of Hypertensive Patients’ EHRs Definition

In our study we assessed the quality of information in hypertensive patients’ EHRs. Quality of information was estimated only by completeness and consistency of information about risk factors (age and gender are not included as they are automatically entered in demographic part of case history), DM and subclinical organ damage. We assessed completeness as compliance of data about risk factors and subclinical organ damage with current guidelines. This means that EHRs contained any information about all required risk factors, subclinical organ damage or both (in the history of the disease, diagnosis or test results). The term consistency was used for evaluation of the logic and the lack of contradiction with current guidelines of data about specific factors. For instance, for dyslipidemia, or DM confirmatory test results were required, or appropriate referrals for analysis recommended in the prescription field.

The risk factors or subclinical organ damage were defined as evaluated if the results of previous examinations were provided in the case history, or when the physician referred the patient to appropriate investigation (for instance, creatinine level for eGFR estimation). DM and risk factors such as dyslipidemia, and glucose tolerance disorder were extracted by automated data mining from the field ‘Diagnosis’. Data about statins therapy was extracted from prescriptions and compared with guideline recommendations.

2.4 EHR Mistakes

For the analysis we used only initial reception records. We analyzed 32158 records (first year: 5862, second year: 5160, third year: 4705; fourth year: 4731, fifth year: 5368, sixth year: 6332). We determined nine classes of the most common mistakes in the assessment of risk factors, DM, subclinical organ damage:
  1. 1.

    Information about dyslipidemia was recorded in the diagnosis field in 49.9% (16037) of primary visits and in 94.2% (15106) it was not confirmed by test results. Moreover, among all patients diagnosed with dyslipidemia, only 58.5% (9389) received statin treatment.

     
  2. 2.

    Information about dyslipidemia was absent (in diagnosis field, previous test results data) in 46.8% (15039) of the patients. Furthermore, 67.1% (10084) of them were not further referred for lipid profile evaluation and in 14.8% (1491) statins were prescribed to patients without disclosing information about indications for lipid-lowering therapy. Assessment of only cholesterol level was recommended in 6.4% (644) of cases.

     
  3. 3.

    In 2.4% (787) of the cases the test results revealed abnormal lipid profiles, but dyslipidemia was not registered in the diagnosis field. Statins were not prescribed to 79.3% (624) of these patients, though they were recommended according to current guidelines.

     
  4. 4.

    Information about glucose tolerance disorder was present in 4.5% (1432) of records, but in 92.8% (1329) of cases was not confirmed by the information about abnormal test results.

     
  5. 5.

    Information about glucose metabolism was absent (in diagnosis field, previous tests results data) in 90.2% (29011) of the patients. Moreover, 99.7% (28920) of them were not further referred for glucose and HbA1c evaluation. Assessment of only glucose level and only HbA1c level was recommended in 5.4% (1551) and 1.7% (483) of cases, respectively.

     
  6. 6.

    Information about BMI was present in 87.6% (28179) of records.

     
  7. 7.

    Information about LVH, eGFR and microalbuminuria was absent in 42.4% (13640), 76.5% (24590) and 99.2% (31916) of the patients, respectively. This means that previous test results were not provided in the case history, and referral to appropriate investigation was not recommended.

     
  8. 8.

    Information about DM was present in 15% (4818) of records, but in 94.5% (4555) of cases were not confirmed by the abnormal test results.

     
  9. 9.

    In DM patients, assessment of only glucose level and only HbA1c level was recommended in 5.6% (1446) and 1.3% (380) of cases, respectively.

     

Risk factors such as smoking and family history of premature CVD were evaluated in all patients.

2.5 Statistical Data Analysis

Completeness and consistency of data may increase with time and experience of EHR system usage (learning curves). We assessed 6-year trends in the proportion of patients without complete assessment of risk factors (dyslipidemia, increased glucose level, BMI). Results are presented in Fig. 1.
Fig. 1.

Six-year trends in proportion of patients with incomplete assessment of risk factors

There were no statistically significant changes in the proportion of patients without complete assessment of risk factors from 2010 till 2013. While during 2013-2015 there was a statistically significant at accepted level = 0.05 increase by 108% (R 2  = 0.99; p = 0.05) of cases with incomplete risk factors assessment. This indicates that the quality of patient EHRs estimated by the completeness and consistency of information about risk factors significantly decreased since 2013.

Six-year trends in the proportion of patients with dyslipidemia, but without recommendations for treatment with statins (for lowering of lipids level and improving of prognosis) in the prescription field are presented in Fig. 2.
Fig. 2.

Six-year trends in proportion of patients with dyslipidemia but without recommended statin therapy

During 2010–2013 the proportion of patients with dyslipidemia not on statin therapy was statistically significant at accepted level = 0.05 reduced by 9% (R 2  = 0.97; p = 0.01). While during 2013–2015 there was a statistically significant at accepted level = 0.05 growth by 64% (R 2  = 0.99; p = 0.05). So again, the completeness of data about dyslipidemia and statin therapy significantly decreased since 2013.

We assessed 6-year trends in the proportion of patients without complete assessment of subclinical organ damage. Results are presented in Fig. 3.
Fig. 3.

Six-year trends in proportion of patients with incomplete assessment of subclinical organ damage

From 2010 to 2013 there was a statistically significant at accepted level = 0.05 descent in patients with incomplete assessment of subclinical organ damage by 34% (R 2  = 0.93; p = 0.04). But again, since 2013 there was a statistically significant at accepted level = 0.05 increase by 42% (R 2  = 0.99; p = 0.01). This means that the quality of patient EHRs estimated by the completeness and consistency of information about subclinical organ damage significantly decreased since 2013 compared to 2012–2013 time interval.

We assessed 6-year trends in the proportion of patients with DM without appropriate assessment or confirmatory test data (results of evaluation of fasting glucose level, HbA1c level or referrals to laboratory). Results are presented in Fig. 4.
Fig. 4.

Six-year trends in proportion of diabetes patients without appropriate assessment of glucose metabolism

There were no statistically significant changes in 2010–2013, while during 2013–2015 there was a statistically significant at accepted level = 0.15 reduction by 6% (R 2  = 0.95; p = 0.14). This means that the quality of patient EHRs about DM significantly increased since 2013, which is inconsistent with data about risk factors and subclinical organ damage.

We hypothesized that negative trends may be related to an increase in the number of primary visits. The samples volume was very small and the use of probability laws may lead to a shifted estimate of statistical characteristics. Therefore, we used coefficients of proportionality of elements for estimation of the relationship between completeness of risk factors and subclinical organ damage assessment, and number of primary visits. The mean values of the proportionality coefficients were approximately equivalent for risk factors and subclinical organ damage (k 1  = 1.02; k 2  = 1.18; Δ = ±0.16 and k 1  = 1.02; k 2  = 1.00; Δ = ±0.02, respectively). Results demonstrated a direct relationship between the studied parameters. This means that the proportion of patients without complete assessment of risk factors and subclinical organ damage increased in parallel with the workload of the clinic (estimated by the number of primary visits due to uncontrolled hypertension).

3 Discussion and Future Works

Results demonstrated that the EHRs without CDSS provides only incomplete and inconsistent information on risk factors, subclinical organ damage and DM. Interestingly, during first three years after implementation of EHRs there was increase in the completeness and consistency of entered data, indicating the existence of learning curves with experience of usage. While after 2013 there was a dramatic progressive drop in the completeness and consistence of stored data. One of the possible explanations is the increase in workload. This theory is confirmed by direct relationships between quality of assessment of risk factors and subclinical organ damage, and number of primary visits. Although we failed to find more factors explaining this phenomenon, we will continue our efforts in defining other regulatory, social or economic reasons for such dramatic deterioration of quality of EHR information after 2013. We will continue research on the quality of hypertensive patients EHRs data on treatment and drug therapy. It is of interest to establish whether the negative trends after 2013 will also affect the quality of drug treatment data.

The collection of detailed clinical information about reported cases, which is necessary for confirmation of the diagnosis and determinations of disease-related risk factors and subclinical organ damage, is still heavily dependent on manual processes implemented by a physician. In order to improve of the quality of data entered in EHRs, we are developing CDSS. This CDSS will be integrated with EHRs system. The CDSS will be able to integrate the heterogeneous data sources implementing semantic description and unification of reporting data (data/information fusion). The architecture of data-driven CDSS was developed on the basis of the approach proposed earlier in [10]. The main three parts of the CDSS are decisions support module, rule base, and knowledge base. Decisions support module, which will be the core of the CDSS, will provide the functional of pre-processing, filtering, and expansion data, stored in EHRs. These functions provides assessment completeness and consistent of information by means of rule base. Rule base will be filled on the basis of current hypertension guidelines. In addition, decisions support module should provide formation of knowledge base, focused on completeness and consistent of clinical data, and medical records similar precedents. Using EHRs for clinical research requires the completeness and consistency of data in EHRs, so a properly designed CDSS may be a key success factor.

4 Conclusion

In summary, results demonstrated that the quality of information about risk factors, subclinical organ damage and DM, stored hypertensive patients’ EHRs, is mostly incomplete and inconsistent. Six-year trends revealed that the quality of hypertensive patients’ EHRs has a tendency to deteriorate with an increase in workload (estimated by the number of primary visits). Experience with working in EHRs system has some positive impact, but it disappears with an increase in workload. We assume the CDSS, which we are developing, implementing rule-based, data-driven, knowledge-based, and hybrid approaches, will be able to improve the quality of hypertensive patients’ EHRs in the cardiological medical research centre. The learning curves could be used to validate the quality of hypertensive patients’ EHRs after the introduction of CDSS.

Notes

Acknowledgments

This paper is financially supported by The Russian Scientific Foundation, Agreement #14-11-00826 (10.07.2014).

References

  1. 1.
    Birkhead, G.S., et al.: Uses of electronic health records for public health surveillance to advance public health. Annu. Rev. Public Health 36, 345–359 (2015). doi: 10.1146/annurev-publhealth-031914-122747 CrossRefGoogle Scholar
  2. 2.
    Coorevits, P., Sundgren, M., Klein, G.O., Bahr, A., Claerhout, B., Daniel, C., et al.: Electronic health records: new opportunities for clinical research. J. Intern. Med. 274(6), 547–560 (2013). doi: 10.1111/joim.12119 CrossRefGoogle Scholar
  3. 3.
    Cruz-Correia, R., Rodrigues, P., Freitas, A., Almeida, F., Chen, R., Costa-Pereira, A.: Data quality and integration issues in electronic health records. In: Hristidis, V. (ed.) Information Discovery on Electronic Health Records, pp. 55–95. CRC Press, London (2009)Google Scholar
  4. 4.
    De Moor, G., et al.: Using electronic health records for clinical research: the case of the EHR4CR project. J. Biomed. Inform. 53, 162–173 (2015). doi: 10.1016/j.jbi.2014.10.006 CrossRefGoogle Scholar
  5. 5.
    Dugas, M., Lange, M., Muller-Tidow, C., Kirchhof, P., Prokosch, H.-U.: Routine data from hospital information systems can support patient recruitment for clinical studies. Clin. Trials 7(2), 183–189 (2010). doi: 10.1177/1740774510363013 CrossRefGoogle Scholar
  6. 6.
    Embi, P.J., Jain, A., Clark, J., Harris, C.M.: Development of an electronic health record-based clinical trial alert system to enhance recruitment at the point of care. In: AMIA Annual Symposium Proceedings, pp. 231–235 (2005)Google Scholar
  7. 7.
    Geissbuhler, A., Safran, C., Buchan, I., Bellazzi, R., Labkoff, S., Eilenberg, K., et al.: Trustworthy reuse of health data: a transnational perspective. Int. J. Med. Inform. 82(1), 1–9 (2013). doi: 10.1016/j.ijmedinf.2012.11.003 CrossRefGoogle Scholar
  8. 8.
    Holzer, K., Gall, W.: Utilizing IHE-based electronic health record systems for secondary use. Meth. Inf. Med. 50(4), 319–325 (2011). doi: 10.3414/ME10-01-0060 CrossRefGoogle Scholar
  9. 9.
    Jensen, P.B., Jensen, L.J., Brunak, S.: Mining electronic health records: towards better research applications and clinical care. Nat. Rev. Genet. 13(6), 395–405 (2012). doi: 10.1038/nrg3208 CrossRefGoogle Scholar
  10. 10.
    Kovalchuk, S.V., Knyazkov, K.V., Syomov, I.I., Yakovlev, A.N., Boukhanovsky, A.V.: Personalized clinical decision support with complex hospital-level modelling. Procedia Comput. Sci. 66, 392–401 (2015). doi: 10.1016/j.procs.2015.11.045 CrossRefGoogle Scholar
  11. 11.
    Mancia, G., et al.: 2013 ESH/ESC guidelines for the management of arterial hypertension: the task force for the management of arterial hypertension of the European Society of Hypertension (ESH) and of the European Society of Cardiology (ESC). Eur. Heart J. 34(28), 2159–2219 (2013). doi: 10.1093/eurheartj/eht151 CrossRefGoogle Scholar
  12. 12.
    Prokosch, H.-U., Ganslandt, T.: Perspectives for medical informatics - reusing the electronic medical record for clinical research. Meth. Inf. Med. 48(1), 38–44 (2009). doi: 10.3414/ME9132 Google Scholar
  13. 13.
    Sittig, D.F., Singh, H.: Defining health information technology-related errors: new developments since to err is human. Arch. Intern. Med. 171(14), 1281–1284 (2011). doi: 10.1001/archinternmed.2011.327 CrossRefGoogle Scholar
  14. 14.
    Sundgren, M., Wilson, P., De Zegher, I.: Making the most of the electronic age. Eur. Pharm. Contractor 3, 18–21 (2009)Google Scholar
  15. 15.
    Turisco, F., Keogh, D., Stubbs, C., Glaser, J., Crowley Jr., W.F.: Current status of integrating information technologies into the clinical research enterprise within US Academic Health Centers: Strategic value and opportunities for investment. J. Investig. Med. 53(8), 425–433 (2005). doi: 10.2310/6650.2005.53806 CrossRefGoogle Scholar
  16. 16.
    Weiskopf, N.G., Weng, C.: Methods and dimensions of electronic health record data quality assessment: Enabling reuse for clinical research. J. Am. Med. Inform. Assoc. 20(1), 144–151 (2013). doi: 10.1136/amiajnl-2011-000681 CrossRefGoogle Scholar
  17. 17.
    Zhou, L., et al.: The relationship between electronic health record use and quality of care over time. J. Am. Med. Inform. Assoc. 16(4), 457–464 (2009). doi: 10.1197/jamia.M3128 CrossRefGoogle Scholar

Copyright information

© Springer International Publishing AG 2016

Authors and Affiliations

  • Anna Semakova
    • 1
  • Nadezhda Zvartau
    • 2
  • Ekaterina Bolgova
    • 1
  • Aleksandra Konradi
    • 2
  1. 1.ITMO UniversitySaint PetersburgRussia
  2. 2.Almazov Federal North-West Medical Research CenterSaint PetersburgRussia

Personalised recommendations