AUTOMATIC SPEECH RECOGNITION AND CLINICAL DOCUMENTATION TIME IN TWO HEALTHCARE INSTITUTIONS IN BUKHARA, UZBEKISTAN: A WITHIN-CLINICIAN QUASI-EXPERIMENTAL STUDY

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Clinical documentation can consume a substantial proportion of physicians’ working time and may limit the operational benefits of healthcare digitalisation. This study examined whether automatic speech recognition (ASR) reduced the time required to create medical documentation in two functioning healthcare institutions in Bukhara, Uzbekistan. A within-clinician quasi-experimental comparison was used: the same clinicians completed documentation by conventional keyboard entry and by ASR-supported voice input. The quantitative analysis covered more than 120 observed patient records. To account for differences in note length, documentation time was standardised as minutes required to produce 100 words. Mean time declined from 8.8 minutes with manual typing to 5.2 minutes with ASR, corresponding to a descriptive reduction of 3.6 minutes per 100 words, or 40.9%. The finding suggests that locally adapted speech input may reduce the labour-time cost of documentation in routine clinical practice. However, the available report did not include dispersion measures, confidence intervals, inferential statistics, transcription accuracy, correction time, or full ethical and technical details. The result should therefore be interpreted as preliminary real-world evidence rather than a definitive causal estimate. Future evaluation should combine time outcomes with note quality, clinically significant error rates, clinician workload, patient consent, privacy safeguards, and implementation costs.