Factors associated with inpatient medical record data quality in hospital settings: a systematic review

Abstract

Background

Inpatient medical record data are an important source of information for clinical care, hospital management, clinical coding, and reimbursement. Poor-quality record data may lead to incomplete documentation, coding errors, inaccurate performance assessment, and distorted reimbursement. In this review, data quality mainly refers to the accuracy, completeness, consistency, timeliness, and validity of inpatient medical record data, including front-page or home-page information and diagnosis/procedure coding data. Evidence on factors associated with these dimensions is scattered across studies of clinical documentation, coding practice, information systems, quality-control, and payment reform. This systematic review aims to synthesise the key associated factors and provide evidence for quality improvement in healthcare organisations.

Methods

This review followed the PRISMA 2020 reporting guideline and was registered in PROSPERO (CRD420251244232). PubMed, Scopus, CNKI, and Wanfang Data were searched for English and Chinese studies published from 1 January 2021 to 7 November 2025. Two reviewers independently screened records, extracted data, and appraised methodological quality using Joanna Briggs Institute tools. Because of heterogeneity in designs, outcomes, and measurements, findings were synthesised narratively.

Results

A total of 29 primary empirical studies were included in the narrative synthesis. Most studies were cross-sectional or retrospective hospital audits, and the overall methodological quality was mainly moderate. Six categories of factors associated with inpatient medical record data quality were identified: clinical documentation and physician-related factors (16 studies), coding staff capacity and professional competence (18 studies), institutional and quality-control mechanisms (19 studies), information-system support (14 studies), training and feedback (20 studies), and external policy or reimbursement pressure (7 studies). The evidence suggests that record quality is shaped by interactions between the data-generation process, coding transformation, organisational governance, technological support, and external incentives.

Conclusion

The available evidence suggests that improving inpatient medical record data quality requires coordinated action across clinical documentation, coding capacity, quality-control mechanisms, information systems, training, and the external policy or reimbursement context. However, because much of the evidence comes from cross-sectional studies and is subject to language restrictions and search-field limitations, causal interpretation should remain cautious. This review provides a structured synthesis to inform future empirical studies and hospital quality-improvement work.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12913-026-15333-0.

Background

Inpatient medical records document patients’ diagnoses, procedures, treatment processes, and discharge outcomes. The data generated from inpatient medical records serve as a primary data source for DRG/DIP payment reform, hospital performance assessment, and refined hospital management. In the context of electronic health records and data-driven hospital governance, the quality of these data has become more visible and more consequential for both clinical and administrative decisions [1, 2]. In this review, inpatient medical record data quality is understood as the extent to which record data are accurate, complete, consistent, timely, and valid for clinical coding and secondary use. This definition covers common quality dimensions reported in the included studies, including completeness of front-page or home-page items (the inpatient record summary page used for coding, quality assessment, and hospital management), accuracy of diagnoses and procedures, consistency between clinical information and codes, timeliness of documentation, and ICD coding validity.

Poor-quality inpatient record data can lead to missing diagnoses, inaccurate principal diagnosis selection, procedural coding errors, distorted DRG/DIP grouping, and biased estimates of resource use [2]. These problems may affect hospital management, quality monitoring, reimbursement accuracy, and the credibility of routine health data. The topic is particularly relevant now because many hospitals rely increasingly on electronic records, automated checks, performance dashboards, and payment systems that draw directly from inpatient record and coding data.

Previous work has examined related topics, including a systematic review of electronic record-driven quality measurement and feedback systems and an empirical programme to improve electronic medical record documentation [3, 4]. However, individual primary studies often focus on one narrow factor, such as clinician documentation behaviour, coder competence, EMR logic checks, hospital quality-control processes, or DRG/DIP payment pressure. As a result, evidence remains fragmented across data-quality dimensions and organisational levels. A synthesis focused on factors associated with inpatient medical record data quality can clarify how these factors are connected and where hospital interventions may be directed.

This systematic review therefore aims to identify, classify, and synthesise factors associated with inpatient medical record data quality in hospital settings, with attention to clinical, coding, organisational, technological, training, and policy-related levels.

Methods

Design and registration

This systematic review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020) reporting guideline. The review was registered with the International Prospective Register of Systematic Reviews (PROSPERO) under registration number CRD420251244232. A completed PRISMA 2020 checklist has been prepared for separate submission.

Information sources and search strategy

We searched PubMed, Scopus, CNKI, and Wanfang Data. PubMed and Scopus were used to capture international biomedical, health information management, and hospital management literature. CNKI and Wanfang Data were included because many studies on inpatient case home-page/front-page quality, ICD coding, DRG/DIP payment reform, and hospital medical record quality-control practice are published in Chinese-language journals and are not consistently indexed in PubMed or Scopus.

The search covered studies published from 1 January 2021 to 7 November 2025, with the final database search conducted on 7 November 2025. This period was selected because recent years have seen rapid expansion of electronic medical record use, DRG/DIP payment reforms, and data-driven hospital performance management in the included literature [1, 2]. The search was limited to English and Chinese publications because these were the languages that the review team could assess reliably.

Search terms covered medical records, electronic medical records/electronic health records, ICD, coding, quality, accuracy, completeness, consistency, and management. ICD and coding terms were included because diagnosis and procedure coding validity is a core dimension of inpatient record data quality and is closely linked to DRG/DIP grouping [1, 2]. Management terms were included because hospital quality-control mechanisms, audit systems, training, and institutional interventions were expected to be relevant to the review question.

Search fields were adapted to each database. Searches were focused on title and keyword-related fields where available, and the full Scopus search strategy, including Boolean operators, search fields, limits, final search date, and the number of records retrieved on that date, is provided in Supplementary File 1. Searches were limited to title and keyword fields after consultation with a librarian, because broad terms such as “electronic medical record” and “quality” retrieved many irrelevant records when abstract fields were included. However, titles and available abstracts were independently screened by two reviewers after duplicate removal. This search-field restriction was considered when interpreting the findings.

Eligibility criteria

Studies were included if they met all of the following criteria: (1) primary empirical studies, including cross-sectional, retrospective, qualitative, or quasi-experimental/before-and-after designs; (2) studies published in English or Chinese; (3) studies focused on inpatient medical record data, front-page/home-page data, electronic medical record documentation data, or diagnosis/procedure coding data; and (4) studies that examined factors associated with data quality or evaluated quality-control interventions relevant to data quality.

Studies were excluded if they were not available in English or Chinese, had no accessible full text, did not focus on inpatient medical record data quality, did not report relevant factors or quality-control measures, lacked a clear hospital setting or data source, were not focused on medical record data quality, or were judged to have serious methodological limitations and high risk of bias after topic and full-text eligibility assessment. Reviews, commentaries, editorials, and conference abstracts without full text were excluded from the final synthesis.

After topic and full-text eligibility were assessed, studies that otherwise met the substantive inclusion criteria were appraised using JBI tools. Studies with serious limitations in sample description, measurement, confounding control, or analytical transparency were considered at high risk of bias and were not included in the final synthesis.

Study selection, data extraction, and quality appraisal

After deduplication in Zotero, two reviewers independently screened titles and abstracts and then assessed potentially eligible full-text reports. Disagreements were resolved through discussion, and unresolved issues were checked by another member of the review team. Data were extracted using a standard form covering author, year, country or region, study setting, study design, sample size, data source, outcome or data-quality dimension, reported factors, methods, and main findings.

Methodological quality appraisal

Methodological quality was assessed using Joanna Briggs Institute (JBI) critical appraisal tools appropriate to study design [5]. Methodological quality appraisal was conducted only after topic relevance and full-text eligibility had been assessed. Reports that otherwise met the eligibility criteria but were judged to have serious methodological limitations and high risk of bias were excluded from the final synthesis, while the quality of the 29 included studies was used to guide interpretation. Cross-sectional and analytical cross-sectional studies were assessed using the JBI checklist for analytical cross-sectional studies, qualitative studies using the JBI qualitative checklist, and quasi-experimental/before-and-after studies using the JBI checklist for quasi-experimental studies. Two reviewers conducted quality appraisal independently. Study quality was summarised descriptively as moderately low, moderate, moderately high, or close to high. In the synthesis, findings supported by multiple moderate or higher-quality studies were given greater interpretive weight than findings from isolated or moderately low-quality studies. The detailed item-level JBI quality appraisal table is provided in Supplementary File 2.

Data synthesis

Because the included studies differed substantially in design, data-quality definitions, outcome measures, and reported statistics, meta-analysis was not appropriate. Subgroup analysis by country, hospital grade, or quality dimension was also not performed because many studies did not report comparable denominators or standardised outcome metrics. A narrative synthesis approach was therefore used to summarise patterns across studies and to identify recurring categories of factors associated with inpatient medical record data quality [5].

The synthesis proceeded in five steps. First, extracted findings were tabulated by study design, quality dimension, and reported factor. Second, similar concepts were coded together, for example, incomplete documentation, delayed completion, ICD coding errors, logical verification, audit feedback, and training. Third, codes were compared across studies and grouped into broader categories. Fourth, the number and quality of studies supporting each category were checked against the evidence matrix. Fifth, the categories were organised into a multi-level framework linking clinical data generation, coding transformation, organisational governance, technological support, training, and policy pressure.

Results

Study selection and characteristics of included studies

The selection process is shown in Fig. 1. The database search identified 1,475 records. After 121 duplicates were removed, 1,354 records were screened. A total of 912 records were excluded after title and, where available, abstract screening. Of 442 reports sought for retrieval, 3 were not retrieved. The remaining 439 reports were further screened based on title, available abstract, keywords, and relevance to the review question. Reports were excluded because they did not analyse factors associated with medical record quality (n = 300), were not relevant to the core topic (n = 49), were not focused on medical record data quality (n = 56), had high risk of bias (n = 4), or were secondary reviews excluded to avoid duplication of evidence (n = 1). Finally, 29 primary empirical studies, corresponding to 29 reports, were included in the narrative synthesis.

Fig. 1.

PRISMA flow diagram of study selection

The included studies were mainly hospital-based studies of inpatient record front-page/home-page data, EMR documentation, diagnosis/procedure coding quality, and quality-control interventions. Study designs included cross-sectional audits, retrospective analyses, qualitative research, and before-after quality-improvement studies. Sample sizes ranged from several hundred to tens of thousands of medical records or related observations.

Records identified: PubMed (n = 211), Scopus (n = 186), CNKI (n = 23), and Wanfang Data (n = 1055); duplicates removed (n = 121); records screened (n = 1354); records excluded after title and, where available, abstract screening (n = 912); reports sought for retrieval (n = 442); reports not retrieved (n = 3); reports further screened for relevance to the review question (n = 439); reports excluded: no analysis of factors associated with medical record quality (n = 300), topic not relevant to the core review question (n = 49), not focused on medical record data quality (n = 56), high risk of bias (n = 4), secondary review excluded to avoid duplication of evidence (n = 1); studies included in review (n = 29); reports of included studies (n = 29).

Methodological quality of included studies

The methodological quality of the 29 included primary studies was mainly moderate. Nineteen studies were rated as moderate quality [1, 4, 6–22], seven as moderately high [2, 23–28], one as close to high [29], and two as moderately low [30, 31]. Common limitations included incomplete control for confounding factors, inconsistent measurement of data-quality dimensions, and the predominance of single-centre or retrospective designs. No randomised controlled trial was identified. A concise summary of methodological quality is provided in Table 1, and detailed item-level appraisal is provided in Supplementary File 2 to improve readability.

Table 1.

Summary of included studies and methodological quality

Characteristic	Summary
Number of included studies	29 primary empirical reports
Country/region	China (n = 25), Iran (n = 2), Indonesia (n = 1), and Australia (n = 1).
Data source and setting	Hospital/inpatient medical record front-page or home-page data, EMR documentation data, and diagnosis/procedure coding data.
Dominant study designs	Mainly cross-sectional or retrospective audits; limited qualitative and before-after/quality-improvement designs.
Sample size	Medical record samples ranged from several hundred to tens of thousands; some studies used physicians, coders, or other hospital staff as respondents or informants.
Methodological quality	19 studies were rated as moderate quality, 7 as moderately high, 1 as close to high, and 2 as moderately low. Low-quality/high-risk-of-bias reports were excluded.
Use of quality appraisal in synthesis	Consistent findings from moderate or higher-quality studies were weighted more strongly; moderately low-quality evidence was interpreted as supportive rather than decisive.

Note: Detailed item-level quality appraisal is provided in Supplementary File 2

Quality appraisal was used to inform the interpretation of the synthesis. Findings supported by multiple moderate or higher-quality studies were interpreted with greater confidence than findings from isolated or moderately low-quality reports.

Factors associated with inpatient medical record data quality: narrative synthesis

The narrative synthesis identified six categories of factors associated with inpatient medical record data quality (Table 2). These categories should be interpreted as connected components of a multi-level pathway from clinical data generation to standardised coding, organisational governance, technological support, continuous training, and external policy pressure.

Table 2.

Evidence matrix of factors associated with inpatient medical record data quality

Factor category	Key points	Supporting studies	Common designs	Quality	Interpretation
Clinical documentation and physician-related factors	Documentation awareness, habits, workload, timeliness, completeness, and understanding of required items.	16 studies [6, 8–10, 12, 16–20, 23, 26–30]	Cross-sectional or retrospective audits	Mostly moderate	Clinicians are the source of many data elements; documentation behaviour affects completeness and accuracy.
Coding staff capacity and professional competence	ICD proficiency, principal diagnosis selection, procedure coding, coder staffing, and coder-clinician communication.	18 studies[2, 6, 10, 11, 13–21, 23, 26–29]	Audits and qualitative studies	Mostly moderate	Coding translates clinical text into standardised data for grouping, statistics, and reimbursement.
Institutional and quality-control mechanisms	PDCA, multi-level audit, rules, feedback, supervision, and performance appraisal.	19 studies[2, 4, 8–12, 14, 15, 17, 19, 22, 24–29, 31]	Cross-sectional and QI designs	Moderate to high	Governance mechanisms support stable data quality, but weak feedback or superficial enforcement may reduce effect.
Information-system support	Structured fields, logical checks, mandatory items, prompts, and HIS/LIS/EMR interconnection.	14 studies [2, 4, 6, 14, 16–22, 24, 25, 27]	Cross-sectional or retrospective audits	Mostly moderate	Systems can reduce omissions and consistency errors, but depend on design and user compliance.
Training and feedback	Regular stratified training, case-based instruction, audit feedback, retraining, and multidisciplinary communication.	20 studies [2, 8, 10–12, 14–22, 24–29]	Cross-sectional and QI designs	Mostly moderate	Training links identified defects to behaviour change and process improvement.
External policy and reimbursement pressure	DRG/DIP payment, performance appraisal, insurance review, and reimbursement incentives.	7 studies[1, 15, 16, 21, 26, 27, 31]	Cross-sectional and QI designs	Moderate	External pressure can raise managerial attention but may create strategic documentation or upcoding risks.

Clinical documentation and physician-related factors

Sixteen studies reported physician-related factors associated with data quality [6, 8–10, 12, 16–20, 23, 26–30]. Common problems included incomplete documentation, delayed completion of record items, inconsistent writing habits, insufficient awareness of the importance of record data, and workload-related omissions. These issues were particularly visible in front-page or home-page fields, principal diagnosis documentation, admission route information, and procedure-related items. The evidence indicates that clinical documentation behaviour forms the first step in the medical record data-quality chain.

Coding staff capacity and professional competence

Eighteen studies identified coding capacity as an important factor associated with data quality [2, 6, 10, 11, 13–21, 23, 26–29]. Coders’ understanding of ICD rules, ability to interpret clinical diagnoses, selection of the principal diagnosis, recognition of comorbidities and complications, and procedure-coding skills influenced whether clinical information was translated accurately into standardised data. Communication between clinicians and coders was also important because ambiguous clinical terminology may lead to coding errors. Staffing issues such as coder shortage were treated under this category because they affect review capacity and the practical performance of coding work.

Institutional and quality-control mechanisms

Nineteen studies supported the role of institutional and quality-control mechanisms [2, 4, 8–12, 14, 15, 17, 19, 22, 24–29, 31]. Reported mechanisms included PDCA cycle management, departmental and hospital-level audits, three-level review systems, quality-control indicators, and links between record quality and performance appraisal. These mechanisms were associated with fewer defects when they provided timely feedback and clear accountability. However, evidence also suggested that delayed feedback, weak enforcement, or formalistic review processes reduced their value.

Information-system support

Fourteen studies reported information-system-related factors [2, 4, 6, 14, 16–22, 24, 25, 27]. Structured fields, mandatory completion rules, logical checks, automatic prompts, and interconnection among EMR, HIS, LIS, and coding systems were commonly described as tools to reduce avoidable errors. These functions may prevent omissions, inconsistent demographic or diagnosis information, and common coding-related mistakes. Nevertheless, information systems should be interpreted as enabling conditions rather than automatic solutions, because data quality still depends on clinical entry, coding judgement, and governance.

Training and feedback

Twenty studies identified training and feedback as important operational mechanisms [2, 8, 10–12, 14–22, 24–29]. One-off training was generally considered insufficient. More consistent evidence supported regular, stratified, case-based training for physicians and coders, combined with audit feedback and retraining on recurring defects. The loop of training, audit, feedback, and retraining links quality-control findings to staff learning and practical behaviour change.

External policy and reimbursement pressure

Seven studies discussed external policy and reimbursement pressure [1, 15, 16, 21, 26, 27, 31]. DRG/DIP payment reform and public hospital performance appraisal increased hospital attention to front-page data, coding accuracy, and standardised record management. These incentives can support investment in coding staff, information systems, and quality-control processes. At the same time, the literature cautions that payment-related pressure should be accompanied by authenticity checks to avoid strategic documentation, upcoding, or other distortions of record data.

Multi-level relationship among the factor categories

Together, the six categories suggest a multi-level quality pathway rather than six separate lists of factors. At the source layer, clinicians generate the diagnostic, procedural, and discharge information that forms the basis of the inpatient record. At the transformation layer, coders translate clinical information into standardised ICD and DRG/DIP-related data. At the process layer, institutional quality-control systems and training mechanisms detect, feed back, and correct errors. At the technical layer, information systems provide structured entry, logical checks, prompts, and data interconnection. At the environmental layer, payment reform and performance appraisal increase managerial attention to record quality but may also create risks of strategic documentation if authenticity safeguards are weak.

Discussion

Principal findings

This systematic review synthesised evidence from 29 primary empirical studies and identified six categories of factors associated with inpatient medical record data quality. The results suggest that data quality is shaped by interacting clinical, coding, organisational, technological, training, and policy-related factors rather than by a single isolated cause. Coding capacity appears to be an important transformation point between clinical documentation and standardised administrative data. However, because most included studies were observational or retrospective, coding capacity should be interpreted as a factor associated with data quality rather than as a confirmed causal mediator.

A previous systematic review focused more narrowly on electronic record-driven quality measurement and feedback systems [3]. In contrast, the present review excluded secondary reviews from the final synthesis and focused on primary empirical studies related to inpatient medical record data quality, including front-page/home-page quality, coding accuracy, quality-control mechanisms, information-system support, training, and policy-related factors. Evidence from an empirical documentation improvement programme also supports the need to connect documentation practice with quality monitoring [4].

The evidence is broadly consistent with individual studies that report documentation defects, coding errors, and quality-control weaknesses as common sources of medical record data problems [6, 9, 13, 21]. It also supports studies showing that information-system functions and quality-control programmes may reduce preventable errors when combined with staff training and feedback [2, 4, 22]. At the same time, external payment reform should be interpreted cautiously. DRG/DIP and performance appraisal may increase managerial attention to data quality, but the same incentives can also create risks of strategic documentation or coding distortion if authenticity safeguards are weak [16, 27].

Practical implications

For hospital managers, the findings suggest that single interventions are unlikely to be sufficient. Improving inpatient medical record data quality may require coordinated action across documentation standards, coder training, clinician-coder communication, routine audit and feedback, and information-system validation functions. These recommendations should be interpreted as evidence-informed directions rather than definitive intervention effects.

At the organisational level, PDCA-based quality-control, timely audit feedback, and clear accountability can support continuous improvement [22, 27]. At the technical level, structured fields, logic checks, and interoperability may reduce avoidable omissions and inconsistencies [2, 4, 21]. At the staff level, regular training for clinicians and coders may help reduce recurring errors and improve shared understanding of documentation and coding rules [9, 23].

Limitations and future directions

This review has several limitations. Most included studies used cross-sectional or retrospective designs, which limits causal inference. The search was restricted to English- and Chinese-language publications and did not include abstract fields, which may have reduced search sensitivity. The included studies were heterogeneous in design, setting, definitions, and outcome measures, preventing meta-analysis. Many studies were also from China, which may limit generalisability. Future research should use stronger study designs and more standardised measures of inpatient medical record data quality.

Conclusions

This systematic review suggests that inpatient medical record data quality in hospital settings is associated with six interacting factor categories: clinical documentation, coding capacity, institutional and quality-control mechanisms, information-system support, training and feedback, and external policy or reimbursement pressure. The review contributes by organising dispersed evidence into a structured framework that may inform future empirical testing and hospital quality-improvement work.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ICD

International Classification of Diseases

DRG

Diagnosis-Related Groups

DIP

Diagnosis-Intervention Packet

PDCA

Plan-Do-Check-Act

EMR

Electronic Medical Record

EHR

Electronic Health Record

HIS

Hospital Information System

LIS

Laboratory Information System

Author contributions

LPW conceived the study, conducted the literature review, analysed and interpreted the data, and drafted and revised the manuscript. JWT participated throughout the study, contributed to the study design, literature screening, data extraction, and manuscript revision. LGC and HX independently cross-checked the study selection, data extraction, and quality appraisal of the included studies. ZBMT and KRBM supervised the methodological design and provided guidance on the research framework and interpretation of the findings. All authors read and approved the final manuscript.

Funding

This work was supported by the Science and Technology Bureau of Nanchong City, Sichuan Province, China (project number 23JCYJPT0032).

Data availability

All data analysed during this study are included in this published article and its supplementary information files. Data were derived from previously published studies included in this systematic review.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.