BACKGROUND: It is common to find a high variability in the accuracy of heart failure (HF) diagnosis in electronic primary care medical records (EMR). Our aims were to ascertain (i) whether the prognosis of HF labelled patients whose ejection fraction (EF) was missing in their EMR differed from those that had it registered, and (ii) the causes contributing to the differences in the availability of EF in EMR. METHODS: Retrospective cohort analyses based on clinical records of HF and attended at 52 primary healthcare centres of Barcelona (Spain). Information of 8376 HF patients aged > 40 years followed during five years was analyzed. RESULTS: EF was available only in 8.5% of primary care medical records. Cumulate incidence for mortality and hospitalization from 1st January 2009 to 31th December 2012 was 37.6%. The highest rate was found in patients with missing EF (HR 1.84, 95% CI 1.68 -1.95) compared to those with preserved EF. Patients hospitalized the previous year and those requiring home healthcare (HR 1.81, 95% Confidence Interval 1.68-1.95 and HR 1.58, 95% CI 1.46-1.71, respectively) presented a higher risk of having an adverse outcome. Older patients, those more socio-economically disadvantaged, obese, requiring home healthcare, and taking loop diuretics were less likely to have an EF registered. CONCLUSIONS: EF is poorly recorded in primary care. HF patients with EF missing at medical records had the worst prognosis. They tended to be older, socio-economically disadvantaged, and more fragile.