How to clean a healthcare data using Python

Some of the specific issues in this dataset include:

• Leading and trailing spaces in string columns.
• Non-numeric values in numeric columns (e.g., ‘forty’ instead of 40 in the Age column).
• Incorrect or missing entries in columns like Blood Pressure, Cholesterol, and Visit Date.

The dataset can be downloaded using this GitHub link.

Step-by-step data cleaning process

Let’s dive into the steps to clean this dataset. Below, we break down each step with corresponding Python code to transform the dataset into a clean and analysis-ready format.

Step 1: Load the dataset

First, we need to load the dataset using Pandas:

import pandas as pd 
 
 # Load the messy data 
 df_healthcare_messy = pd.read_csv('path_to/healthcare_messy_data.csv')

NB: Please replace the “path_to” with the actual location of the stored messy data on your device.

Step 2: Strip leading and trailing spaces

This dataset may likely contain unnecessary spaces in string columns, which can cause problems during analysis. We remove these spaces using the following code:

# Strip leading and trailing spaces from string columns 
 for column in df_healthcare_messy.select_dtypes(include=['object']).columns: 
 df_healthcare_messy[column] = df_healthcare_messy[column].str.strip()

Step 3: Correct non-numeric values

Some columns, like Age, contain non-numeric values that should be converted:

# Correct the Age column 
df_healthcare_messy['Age'] = df_healthcare_messy['Age'].replace('forty', 40)  # Replace 'forty' with 40
df_healthcare_messy['Age'] = pd.to_numeric(df_healthcare_messy['Age'], errors='coerce')  # Convert to numeric, coercing errors to NaN

# Fill NaN values in the Age column before converting to int
df_healthcare_messy['Age'] = df_healthcare_messy['Age'].fillna(0).round(0).astype(int)  # Remove decimals and convert to int

Similarly, for other numeric columns like Blood Pressure and Cholesterol:

# Correct the Blood Pressure column 
 df_healthcare_messy['Blood Pressure'] = df_healthcare_messy['Blood Pressure'].replace('NaN', np.nan) 
 
 # Correct the Cholesterol column 
df_healthcare_messy['Cholesterol'] = pd.to_numeric(df_healthcare_messy['Cholesterol'], errors='coerce')  # Convert to numeric, coercing errors to NaN

# Fill NaN values in the Cholesterol column before converting to int
df_healthcare_messy['Cholesterol'] = df_healthcare_messy['Cholesterol'].fillna(0).round(0).astype(int)  # Remove decimals and convert to int

Step 4: Standardise date columns

Dates are often stored in various formats. Standardising them ensures consistency:

# Standardise the Visit Date column 
 df_healthcare_messy['Visit Date'] = pd.to_datetime(df_healthcare_messy['Visit Date'], errors='coerce')

Step 5: Handle missing values

Missing data is a common issue in datasets. We can handle missing values by filling them with appropriate defaults:

# Fill NaN values with a default value or strategy 
 df_healthcare_messy['Age'].fillna(df_healthcare_messy['Age'].mean(), inplace=True) 
 df_healthcare_messy['Cholesterol'].fillna(df_healthcare_messy['Cholesterol'].mean(), inplace=True) 
 
 # Fill NaN values for categorical columns with a placeholder or most frequent value 
 df_healthcare_messy['Gender'].fillna('Unknown', inplace=True) 
 df_healthcare_messy['Condition'].fillna(df_healthcare_messy['Condition'].mode()[0], inplace=True) 
 df_healthcare_messy['Medication'].fillna(df_healthcare_messy['Medication'].mode()[0], inplace=True) 
 df_healthcare_messy['Blood Pressure'].fillna('120/80', inplace=True) 
 df_healthcare_messy['Email'].fillna('no_email@example.com', inplace=True) 
 df_healthcare_messy['Phone Number'].fillna('000-000-0000', inplace=True)

Step 6: Save the cleaned dataset

Finally, we save the cleaned dataset to a new CSV file for future use:

# Save the cleaned data to a new CSV file 
 df_healthcare_messy.to_csv('path_to/cleaned_healthcare_messy_data.csv', index=False)

NB: Please replace the “path_to” with the actual location where you would like the cleaned dataset to be stored on your device.

Summary of cleaning steps

To summarise, the following steps were taken to clean the dataset:

• Loaded the dataset using Pandas.
• Stripped leading and trailing spaces from string columns.
• Corrected non-numeric values in numeric columns.
• Standardised date formats in the Visit Date column.
• Handled missing values by filling them with appropriate defaults or most frequent values.
• Saved the cleaned dataset for future analysis.

The cleaned dataset can be download using this GitHub link.

Below is just a snippet of the cleaned dataset:

Patient Name	Age	Gender	Condition	Medication	Visit Date	Blood Pressure	Cholesterol	Email	Phone Number
david lee	25	Other	Heart Disease	METFORMIN	2020-01-15	140/90	200	name@hospital.org	555-555-5555
emily davis	0	Male	Diabetes	NONE		120/80	200	no_email@example.com	000-000-0000
laura martinez	35	Other	Asthma	METFORMIN		110/70	160	contact@domain.com	000-000-0000
michael wilson	0	Male	Diabetes	ALBUTEROL	2020-01-15	110/70	0	name@hospital.org	555-555-5555
david lee	0	Female	Asthma	NONE		110/70	180	no_email@example.com
mary clark	0	Male	Hypertension	METFORMIN		140/90	180	no_email@example.com	000-000-0000
robert brown	0	Male	Hypertension	LISINOPRIL		120/80	0	name@hospital.org	000-000-0000
david lee	60	Other	Asthma	NONE		120/80	0	name@hospital.org	000-000-0000

Conclusion

Cleaning a dataset is an essential step before any data analysis. In this tutorial, we’ve walked through a systematic approach to handle common data issues such as missing values, inconsistent formats, and incorrect data entries in the healthcare data.

With the dataset now clean, you’re ready to perform accurate and meaningful analyses. Remember, a clean dataset is a foundation for reliable insights!

Feel free to adjust the code snippets according to your specific dataset, and happy coding!