Background Falls in older people result in harm for individuals and are a major public health problem, but there is little published data on the recording and incidence of falls seen in primary care, with which to consider the implications of recent policy initiatives. A range of factors contribute to falls risk. Amongst these, the role of some medications is well established, but the evidence base regarding the effects of some of the most commonly prescribed medications remains meagre and inconsistent. Aims The project aims to quantify the overall incidence and distribution of recorded falls among older people in primary care in the UK, and the associated risk of death. The falls risk profile of more recently introduced serotonin noradrenalin reuptake inhibitors (SNRls) is explored to assess whether it is more favourable than that of selective serotonin reuptake inhibitors (SSRls). Similarly, prescribing of subclasses of antihypertensive medication is explored to establish whether any of them modify risk of falling. Finally, other classes or sub-classes of medication prescribed in primary care are identified whose apparent falls risk warrants further investigation. Methods Analysis of falls and prescribing history in the electronic records of patients aged 60 years and over from The Health Improvement Network (THIN) using cohort, survival, case-control and case series study designs. Results Amongst people aged >60 years the overall crude incidence rate of recorded falls was 3.58/100 person-years (3.56-3.61), higher in older age groups, in women and least advantaged social groups, and was constant in the period 2003-2006. Fallers experienced a substantial increase in mortality (two-fold increase for recurrent fallers, and more than five-fold for those aged 60-74 years). This increase is independent of fractures recorded at the time of the fall or subsequently. People who fall have an increased rate of subsequent fracture (approximately three-fold and, for recurrent fallers aged 60-74 years more than eight-fold). There was an increased risk of current prescribing of SNRls (adjusted OR 1.79, 1.42 - 2.25) in first fall cases compared with controls. This was similar in magnitude to that seen with tricyclic antidepressants and SSRls. The increase in risk was apparent within the first 28 days after first prescription. The effects were also apparent in the self-controlled case series analysis: the incidence risk ratio for the period 1-28 days after initiation of treatment compared with unexposed periods was 1.49 (1.15 - 1.93). There was an increased risk of current prescribing of thiazides (adjusted OR 1.28, 1.16-1.42). At 3 weeks after first prescribing the adjusted risk remained 4.28 (1.19-15.42). In the case series analysis the incidence risk ratio for the period 21 days after first prescription was 2.80 (1.7 - 4.57). We found a reduced risk for current prescribing of beta blockers (adjusted OR 0.90; 0.85- 0.96), but a weakly positive effect in the case series analysis for the corresponding period IRR 1.23 (1.02-1.48). Taken together, the case-control and case series analyses of other subclasses of antihypertensives provided weak or no evidence for an effect on falls. In the hypothesis generating case-control analysis of other medication classes, unadjusted odds ratios of greater than 1.7 were found in a number of classes of medication including: laxatives, antifungals, corticosteroids, insulin, antibiotics for mild to moderate acne, and vaccines for influenza and other infections. Conclusions Older people with a recorded fall represent a group who are at increased risk of death, irrespective of whether they have a subsequent fracture. Nevertheless the incidence of falls recorded in primary care suggests that guidance about asking patients if they have fallen in the last year appears not to have been followed during the study period. The fact that the incidence rate of falls is strongly associated with social disadvantage suggests the need to target the design and delivery of interventions accordingly. The falls risk profile of SNRls, which is similar to that of SSRls and TCAs, suggests that clinicians initiating prescribing of SNRls should be alert to the increased risk of falls. Similarly, clinicians initiating prescribing of thiazides in older people, which has generally been considered a 'safe' option for older patients, should be alert to the possibility of an increased risk of falls in the first three weeks of prescribing. Case series analysis of recurrent periodic exposures can elucidate bias in classis case-control analysis of the same data, and will be useful in assessing the falls risk profile of other medications such as insulin. Given the small size of sources of detailed data about older people who fall and the imprecision in their measurement of exposure to medications and potential confounders, case-control and case series analysis of first falls in THIN represents a valuable source of new evidence about medication risk factors.