Discrepancies requiring clarification in cancer patients: a risk predictive model
- Triana González-Carrascosa Vega1,
- Jesús Francisco Sierra-Sánchez2,
- María José Martínez-Bautista1,
- María Victoria Manzano Martín1,
- Jose Manuel Baena Cañada3,
- Irene Romero-Hernández1
- 1Department of Pharmacy, Hospital Universitario Puerta del Mar, Cádiz, Spain
- 2Department of Pharmacy, Hospital Universitario de Fuenlabrada, Fuenlabrada, Spain
- 3Department of Oncology, Hospital Universitario Puerta del Mar, Cádiz, Spain
- Correspondence to Dr Triana González-Carrascosa Vega, Department of Pharmacy, Hospital Universitario Puerta del Mar, C/Ana de Viya, 21, Cádiz CP 11105, Spain;
- Received 23 August 2012
- Revised 18 November 2012
- Accepted 23 November 2012
- Published Online First 20 December 2012
Purpose Medication reconciliation has been proved to be a safe and effective strategy for preventing medication errors. However, the use of this strategy and the need for its implementation in cancer patients has not been adequately studied. This study was performed to develop a predictive model of the risk of discrepancies that need clarification (DNCs) in cancer patients.
Material and methods A retrospective observational study was designed in order to develop a predictive model of the risk of DNCs in cancer patients. Patients diagnosed with breast or colon cancer during 2010 who received chemotherapy and were treated with any chronic medication were included. As endpoints, the incidence of DNCs, drugs involved and risk factors for DNCs were measured. After detecting DNCs, a statistical analysis was performed using multivariate logistic regression with dependent variable discrepancies. The predictive model building was performed using the Hosmer–Lemeshow test.
Results A total of 168 patients were included, 83% of whom were taking home medication. At least one DNC was detected in 50.7% of patients, with a total of 131 DNCs. Multivariate analysis identified two independent variables as predictors of the occurrence of DNCs: the number of cytostatic agents (OR 1.24, 95% 1.03 to 1.49, p=0.026) and the number of target drugs (OR 1.77, 95% CI 1.37 to 2.29, p<0.0001). The proposed predictive model includes two variables to determine the probability (Pr) of suffering a DNC using the following mathematical expression: Pr (%)= 1/(1+e−(−2.21+0.21×Number of cytostatic drugs+0.57×Number of target drugs))×100. The area under the receiver operating characteristic curve was 0.78 (95% CI 0.7 to 0.85, p<0.001), so the model is able to discriminate between patients with and without DNCs with moderate accuracy.
Conclusions Half of all patients who are receiving chemotherapy and are taking home medication have some DNCs. The resulting model can predict the probability of DNCs in cancer patients from the information about their medication.
Patient safety plays an increasingly important role among the quality objectives of health systems. The scale of the problem was revealed in 1999 and 2001 with the publication of two reports by the Institute of Medicine (IOM) in the USA: ‘To err is human: building a safer health system’ and ‘Crossing the quality chasm: a new health system for the 21st century’.
In these reports the estimated annual mortality caused by medical errors in the USA was 44 000 and 98 000 deaths, respectively. Most errors could be attributed to system failures.1 ,2 Following these reports, Health Grades clinical consultants updated the data in a study covering 45% of hospital admissions per year. The results showed that the annual deaths caused by medical errors amounted to 195 000.3
In response to the IOM report, the US federal government issued a comprehensive document on patient safety. This included proposals of actions to carry out the IOM recommendations.4 Similarly, the UK government developed a strategy aimed at reducing care errors by 40%.5
In 2005 the Spanish Government promoted the National Study of Adverse Events related to Hospitalisation (Estudio Nacional sobre los Efectos Adversos ligados a la Hospitalización) and found that 9.3% of patients admitted had an adverse effect derived from hospital care, 37.4% of which were caused by drugs.6 Using these results, the Ministry of Health and Consumption designed a Quality Plan for the National Health System which was published in March 2006. In this plan, areas of action, strategies, objectives and projects aimed to increase the safety of patients were developed.7 Included in these strategies was a set of standards related to different aspects of patient safety. These included a standard process (PR-19) on reconciling home medication. Few studies have shown the potential of this activity to reduce medication errors. A systematic review of 22 studies of medication reconciliation8 shows that, after checking the medical history of 3755 patients, 67% had a medication error.
A randomised clinical trial conducted on surgical patients showed that starting up a reconciliation programme before surgery may decrease the incidence of discrepancies that need clarification (DNCs) by 50%.9 The causes of reconciliation errors are multiple, although Delgado and colleagues10 recognised several factors that are becoming more common: comorbidity and polypharmacy; absence of a single medication record; condition of the patient at the time of hospital admission; and the adaptation of home treatment to the hospital's pharmaceutical guide.
Cancer patients were chosen as the study group in which to implement the improvement of medication safety for several reasons:
Oral and intravenous chemotherapy is on the list of drugs with a high risk of error made by the Institute for Safe Medication Practices.11 This makes patients receiving chemotherapy ideal candidates to help develop and promote programmes that improve the safe use of drugs.
There is a lack of studies about home medication reconciliation in cancer patients. Most studies have been conducted on hospitalised patients or patients who have had surgery.8 However, information on patients receiving chemotherapy admitted for a short time is very limited and in some cases there are few case reports.
The oncology prescription process is frequently done using specific computer applications that are not connected with the patient's chronic clinical record.
The aim of the present study was to identify the risk factors involved in the occurrence of DNCs in patients receiving chemotherapy and to obtain a predictive model of the risk of having DNC in cancer patients receiving chemotherapy.
A retrospective observational study, which included patients with breast cancer or colon cancer diagnosis who began chemotherapy in 2010, was performed in a University Hospital. The inclusion criteria were taking home medication during chemotherapy, a diagnosis of breast or colon cancer and having begun chemotherapy in 2010.
Medication reconciliation is the formal process of obtaining a complete and accurate list of each patient's current home medications including name, dosage, frequency and route of administration, transfer and/or discharge medication. This process therefore aims to make a prescription for the care process considering all the medication the patient has taken up to that point and deciding whether to maintain it as part of the pharmacotherapy in the new clinical situation.
In this study it was assumed that the oncologist did not make any modification to the treatment that was collected in the electronic medical record (e-MR) (named Digital Single Health History—Diraya). Before analysing the patient's treatment, the pharmacotherapeutic history was combined from two information systems that contained the medication records: (1) e-MR was used to obtain information about the patient's home medication prescribed by the primary care physician; and (2) Oncowin was used to obtain information about chemotherapy scheme medication considering two different types of drugs: cytostatics and supportive care medication.
Once the pharmacotherapeutic history was collected, DNCs were analysed between home medication and chemotherapy. As possible improvements in drug therapy, all DNCs found were considered. DNCs are defined as the interactions and duplicities found between home medication and chemotherapy. Drugs that were included in these DNCs and their frequency of occurrence were also identified. To determine the risk factors involved in the development of DNCs, the incidence was first calculated and then a multivariate analysis was conducted. The analysis looked for relationships between the incidence of DNCs and the following independent factors: age, sex, diagnosis, stage, number of cytostatic drugs, number of supportive care drugs, number of home medications, proportion of target drugs and total number of drugs. We considered as target drugs those home treatment drugs that have documented evidence of an interaction with a cytostatic agent or support drugs.
Statistical analysis was performed by multivariate logistic regression with DNCs as the dependent variable. A predictive model of the risk of developing DNCs in cancer patients was also developed. The predictive model building was performed with the Hosmer–Lemeshow test.12 Potential predictors were selected using univariate logistic regression (p<0.2). Multivariate models explored were obtained with inclusion and exclusion criteria of the variables of 0.05 and 0.1, respectively. Statistical analysis was performed using SPSS V.15 (SPSS, Chicago, Illinois, USA). A receiver operating characteristic (ROC) curve was obtained from the proposed model and the area under the curve (AUC) was calculated and used to assess the predictive ability of the model. According to the criteria of Swets,13 the model was assessed by the AUC value obtained (AUC=0.5–0.7, low accuracy predictive model; AUC=0.7–0.9, predictive model that may be useful; and AUC>0.9, high accuracy model. If the 95% CI did not include 0.5 and was above this value, the test was able to discriminate between patients with and without discrepancies.
Of a total of 168 patients who received chemotherapy during the study, 100 (59.5%) had breast cancer and 68 (40.5%) had colon cancer. One hundred and forty (83%) were treated with home medication during chemotherapy, thus being candidates for inclusion in the reconciliation medication process. Patient characteristics are shown in table 1.
After analysing the pharmacotherapeutic profiles, 71 patients (50.7%) had at least one DNC, and a total of 131 DNCs were identified. Table 2 shows the frequency of drugs involved in the DNCs.
The univariate logistic regression analysis excluded some independent variables as predictors, including age, sex, diagnosis and staging. Multivariate analysis identified two independent variables as predictors of the occurrence of DNCs: the number of cytostatic drugs (OR 1.24, 95% CI 1.03 to 1.49, p=0.026) and the number of target drugs (OR 1.77, 95% CI 1.37 to 2.29, p<0.0001). The proposed predictive model includes two variables to determine the probability (Pr) of suffering a DNC by the following mathematical expression: Pr (%)=1/(1+e−(−2.21+0.21×Number of cytostatic drugs+0.57×Number of target drugs))×100. The AUC of the ROC curve (figure 1) was 0.78 (95% CI 0.7 to 0.85, p<0.001). Since the lower limit of the 95% CI was >0.5, we can say that the model is able to discriminate between patients with and without DNCs. Moreover, since the AUC value was between 0.7 and 0.9, the accuracy with which a distinction can be made is intermediate.
Half of the patients with home medication at the time of receiving chemotherapy had at least one DNC. It is known that approximately 80% of the DNCs are confirmed as reconciliation errors.14 The reconciliation medication process provides benefit to patients by identifying and preventing these errors. It can be estimated that one-third of patients receiving chemotherapy and other home medication will obtain a safety benefit from the reconciliation medication process. This approach should be confirmed by the following research, prospectively, including reconciliation errors as the primary endpoint instead of DNC.
Taking account of the fact that the proportion of patients with DNCs found is below the proportion found in other studies,15 ,16 our results could be higher than expected. This is because the most frequent errors found in all studies published are those of omission, accounting for 60–80% of total reconciliation errors.8 ,17 Our study has assumed that patients have continued with their complete home medication treatment. For this reason, there are no omission errors so the estimated proportion of patients in whom the occurrence of reconciliation errors could be prevented is high.
This study has considered all treatments contained in the e-MR, prescribed within the range that included chemotherapy. Other prospective studies have found that up to 70% of the clinical records in the e-MR used (Digital Single Health History—Diraya) contain some inaccuracies, of which excess medication is the most frequent.18 Therefore, although these medicines are part of the patient's medical record, they are not part of their treatment. The extent to which this defect in information sources is present in cancer patients might be different and can be established in studies that include an interview to find out the pharmacotherapeutic history of the cancer patient.
The medications most frequently found to be related to DNCs were ibuprofen, hydrochlorothiazide and tramadol. None of these medicines is directly related to colon or breast cancer treatment. This finding could make this predictive model useful for patients with other types of tumours.
The two variables whose relationship with the risk of DNCs has been identified—number of cytostatic drugs and number of target drugs—can be known before the interview with the patient. This, together with obtaining a predictive model of risk for DNCs in cancer patients, could facilitate prioritisation of care to patients with a higher risk of reconciliation errors.
Half of cancer patients taking any medication while receiving chemotherapy have a DNC. The resulting model can predict the probability of DNCs in cancer patients from the details of their chemotherapy and chronic treatment.
Half of all patients who are receiving chemotherapy and are taking home medication have some DNCs.
The resulting model can predict the probability of DNCs in cancer patients from the information about their medication.
Competing interests None.
Ethics approval Comite De Etica De La Investigacion Del Hospital Universitario ‘Puerta Del Mar’ Y Distrito Bahia De Cadiz La Janda.
Provenance and peer review Not commissioned; externally peer reviewed.