Handling NaN townsend values

[MattStammers]

We are currently losing 330 patients for whom Townsend cannot be calculated. How have you handled this?

[vvcb]

townsend_score_decile.fillna(0) is a workaround. nan is treated as float by Pandas. Modifying the schema to accept float instead of int dtypes to accomodate nan will make values like 3.5 valid as well.

This is documented in the example pipeline.

[MattStammers]

This one can be closed