Heart failure is a long-term condition characterized by the heart being unable to effectively pump blood around the body with it being a common event caused by cardiovascular diseases.

A total of 299 patients diagnosed with heart failure had their information gathered and collated into a data set. The data set included the age, sex, level of creatinine phosphokinase, serum creatinine, serum sodium and platelets in the blood of the patients. It also included the ejection fraction of the patient and whether the patient has hypertension, anaemia and diabetes. The smoking habits of the patients were also recorded. Most importantly the data set highlighted whether the patient died during the follow up period after being diagnosed with heart failure.

The information gathered about the patient is important in helping to predict whether patients diagnosed with heart failure will survive and how long for with the factors acting as prognostic indicators for survival of heart failure.

Age of patients

The age of the patients effect on surviving heart failure are illustrated in the form of a bar plot (Figure 1).
Figure 1. Influence of age of patient has on survival of heart failure

Figure 1. Influence of age of patient has on survival of heart failure

Figure 1 illustrates that the patients which survived heart failure were on average younger then the patients who died. With the mean age of patients which died being 65 while the mean age of patients which survived was 59. This illustrates that age is a prognostic indicator of surviving heart failure with younger patients more likely to survive.

Sex of patients

The influence the sex of the patient has on surviving heart failure are illustrated in the form of a grouped bar chart with the proportion of males and females being used due to unequal size of males and females groups in the data set(Figure 2).
Figure 2. Influence of sex on survival of heart failure illustrated by the proportion of males and females who died or survived

Figure 2. Influence of sex on survival of heart failure illustrated by the proportion of males and females who died or survived

The sex of the patient has minimal influence on the survival of the patient (Figure 2). The proportion of females with heart failure dying being 32.4% while the proportion of males with heart failure dying was 32.0% illustrating only 0.4% difference in proportion of female deaths in comparison to male death. This illustrates that sex cannot be used to predict survival or heart failure.

Smoking habits of patients

The influence the smoking habits of patients has on surviving heart failure are illustrated in the form of a grouped bar chart (Figure 3).
Figure 3. Influence of smoking habits on survival of heart failure illustrated by the proportion of smokers and non-smokers Who died or survived

Figure 3. Influence of smoking habits on survival of heart failure illustrated by the proportion of smokers and non-smokers Who died or survived

Figure 3 shows that smoking can be seen as having no influence on the survival of heart failure in our data with the proportion of Non-smokers who died being 1.3% greater then the proportion of Smokers who died.

Anaemia and Diabetes

The influence Anaemia and diabetes has on patients survival are illustrated in the form of 2 separate stacked bar charts (Figure 4)
Figure 4. Influence of anaemia and diabetes on survival of heart failure

Figure 4. Influence of anaemia and diabetes on survival of heart failure

Anaemia can be seen as influencing the survival of patients with heart faliure. The proportion of patients with anaemia who died is 6.3% greater then the proportion of non anaemic patients who died (figure 4). This illustrates that an individual having anaemia will have less chance of surviving heart failure with anaemia in turn being a prognostic indicator.

Diabetes on the other hand can be seen as having minimal influence on the survival of heart failure. The proportion of patients with diabetes which died was only 0.2% greater then the proportion of non diabetic patients which died. This illustrates diabetes is not an indicator of surviving heart failure.

Hypertension and Ejection factor

The influence hypertension and ejection fraction has on patients surviving heart failure are illustrated by a stacked bar chart for the hypertension and a density plot for the Ejection fraction (Figure 3)
Figure 5. Influence of hypertension and Ejection fraction (%) on survival of heart failure

Figure 5. Influence of hypertension and Ejection fraction (%) on survival of heart failure

Figure 5 demonstrates further factors which influences the survival of heart failure. Figure 5 illustrating that the patients which had hypertension were more likely to die. With 7.7% more patients with hypertension dying in comparison to patients which did not have hypertension.

Figure 5 also demonstrates that the ejection fraction of a patient influences their survival. With patients that died typically having a lower ejection fraction %. With the mean ejection fraction % of patients which died being 33.5% while for patients which survived the mean Ejection fraction % is 40.3%. This illustrates that the ejection fraction of a patient can be used as a prognostic indicator for surviving heart failure.

Biomarker levels

How the levels of 4 different biomarkers changes with heart failure patients who survived and died were compared using violin plots. Violin plots are useful at illustrating the distribution of the data especially for large data sets (n=299).
Figure 6. The level of the biomarkers: Creatine phosphokinase, Serum creatinine, Serum sodium and platelets in patients which survived and died

Figure 6. The level of the biomarkers: Creatine phosphokinase, Serum creatinine, Serum sodium and platelets in patients which survived and died

The 4 violin-plots in figure 6 illustrate how helpful different biomarkers are at indicating whether a patient with heart failure will survive or die. Firstly, figure 6 illustrates that the levels of creatinine phosphokinase, serum sodium and platelets were similar in patients which died and survived. With the violin plots illustrating that for these 3 biomarkers the survived and died violin plots have similar median levels and interquartile ranges (Figure 6). However the levels of Serum creatinine can be seen as varying to a greater extent between the patients which survived and died with the patients which died. With the patients which died having a higher medium serum creatinine level then those who survived. In turn this could suggest that levels of serum creatinine are a good prognostic indicator of survival of heart failure.

However, statistical analysis of figure 6 is required to test if the difference in levels of biomarkers between survived and died is statistically significant.

Statistical analysis

The use of statistical tests will help illustrate any statistical significance between the survival of heart failure patients and the levels of the different biomarkers highlighted in Figure 6.
Table 1 – Biomarker medians and Anderson–Darling test results by sex and survival status
Survival Biomarker N Median A Statistic P-value Significant
Female
Survived Creatinine phosphokinase 71 291.00 5.3771083 < 0.001 Yes
Died Creatinine phosphokinase 34 242.00 5.5410715 < 0.001 Yes
Survived Platelets 71 270000.00 2.9336876 < 0.001 Yes
Died Platelets 34 259179.02 0.4441143 0.269 No
Survived Serum creatinine 71 1.00 8.8146952 < 0.001 Yes
Died Serum creatinine 34 1.55 4.4275311 < 0.001 Yes
Survived Serum sodium 71 137.00 0.7982731 0.037 Yes
Died Serum sodium 34 136.00 0.5673951 0.131 No
Male
Survived Creatinine phosphokinase 132 231.50 16.1220227 < 0.001 Yes
Died Creatinine phosphokinase 62 367.50 12.5603446 < 0.001 Yes
Survived Platelets 132 251000.00 2.5173107 < 0.001 Yes
Died Platelets 62 258000.00 0.6508940 0.085 No
Survived Serum creatinine 132 1.10 13.4245446 < 0.001 Yes
Died Serum creatinine 62 1.30 6.6933679 < 0.001 Yes
Survived Serum sodium 132 137.00 2.0538163 < 0.001 Yes
Died Serum sodium 62 135.00 0.5804684 0.126 No

The use of the Anderson-Darling test (Table 1) helped illustrated the distribution and normality of the data with majority of the data being non-normally distributed (The A statistic was significant). With only the platelets level for female and male patients which died, serum sodium levels in female and male patients which died not being non normally distributed.

In turn with the majority of the data being non normally distributed a non parametric statistical test should be used to analyse if there is a significant difference in the level of the biomarkers between patients which survived and died.

Table 2- Wilcoxon Rank-Sum Test Results
Biomarker W Statistic p-value Significant
Female
Creatinine phosphokinase 1186.5 0.891 No
Platelets 1407.0 0.172 No
Serum creatinine 496.0 < 0.001 Yes
Serum sodium 1394.0 0.2 No
Male
Creatinine phosphokinase 3972.5 0.744 No
Platelets 4064.0 0.94 No
Serum creatinine 2517.0 < 0.001 Yes
Serum sodium 5368.5 < 0.001 Yes

The non parametric test of choice to analyse the levels of biomarkers against the survival of patients is the Wilcoxon Rank-sum test (Table 2). The Wilcoxon rank-sum test was chosen due to it allowing for the comparison of the means of two related groups such as patients which survived and died from heart failure.

The results of the Wilcoxon rank-sum test (Table 2) illustrate tha the levels of serum creatinine in both male and female patients was significantly greater in patients which died then those which survived. In the male patients the level of serum sodium can also be seen as significantly less in patients which survived in comparison to patients which died (Table 2).

The results illustrate that the biomarker serum creatinine can be used as a prognostic indicator for the survival of patients with heart failure.